llm - 2025_04
Navigation
Papers
Large Language Models (LLMs) demonstrate remarkable capabilities in leveraging comprehensive world knowledge and sophisticated reasoning mechanisms for recommendation tasks. However, a notable limitation lies in their inability to effectively model sparse identifiers (e.g., user and item IDs), unlike conventional collaborative filtering models (Collabs.), thus hindering LLM to learn distinctive user-item representations and creating a performance bottleneck. Prior studies indicate that integrating collaborative knowledge from Collabs. into LLMs can mitigate the above limitations and enhance their recommendation performance. Nevertheless, the significant discrepancy in knowledge distribution and semantic space between LLMs and Collab. presents substantial challenges for effective knowledge transfer. To tackle these challenges, we propose a novel framework, SeLLa-Rec, which focuses on achieving alignment between the semantic spaces of Collabs. and LLMs. This alignment fosters effective knowledge fusion, mitigating the influence of discriminative noise and facilitating the deep integration of knowledge from diverse models. Specifically, three special tokens with collaborative knowledge are embedded into the LLM's semantic space through a hybrid projection layer and integrated into task-specific prompts to guide the recommendation process. Experiments conducted on two public benchmark datasets (MovieLens-1M and Amazon Book) demonstrate that SeLLa-Rec achieves state-of-the-art performance.
The dominant metaphor of LLMs-as-minds leads to misleading conceptions of machine agency and is limited in its ability to help both users and developers build the right degree of trust and understanding for outputs from LLMs. It makes it harder to disentangle hallucinations from useful model interactions. This position paper argues that there are fundamental similarities between visual perception and the way LLMs process and present language. These similarities inspire a metaphor for LLMs which could open new avenues for research into interaction paradigms and shared representations. Our visual system metaphor introduces possibilities for addressing these challenges by understanding the information landscape assimilated by LLMs. In this paper we motivate our proposal, introduce the interrelating theories from the fields that inspired this view and discuss research directions that stem from this abstraction.
Large Language Models (LLMs) have emerged as powerful tools for automating various programming tasks, including security-related ones, such as detecting and fixing vulnerabilities. Despite their promising capabilities, when required to produce or modify pre-existing code, LLMs could introduce vulnerabilities unbeknown to the programmer. When analyzing code, they could miss clear vulnerabilities or signal nonexistent ones. In this Systematic Literature Review (SLR), we aim to investigate both the security benefits and potential drawbacks of using LLMs for a variety of code-related tasks. In particular, first we focus on the types of vulnerabilities that could be introduced by LLMs, when used for producing code. Second, we analyze the capabilities of LLMs to detect and fix vulnerabilities, in any given code, and how the prompting strategy of choice impacts their performance in these two tasks. Last, we provide an in-depth analysis on how data poisoning attacks on LLMs can impact performance in the aforementioned tasks.
Hallucination, i.e., generating factually incorrect content, remains a critical challenge for large language models (LLMs). We introduce TOHA, a TOpology-based HAllucination detector in the RAG setting, which leverages a topological divergence metric to quantify the structural properties of graphs induced by attention matrices. Examining the topological divergence between prompt and response subgraphs reveals consistent patterns: higher divergence values in specific attention heads correlate with hallucinated outputs, independent of the dataset. Extensive experiments, including evaluation on question answering and data-to-text tasks, show that our approach achieves state-of-the-art or competitive results on several benchmarks, two of which were annotated by us and are being publicly released to facilitate further research. Beyond its strong in-domain performance, TOHA maintains remarkable domain transferability across multiple open-source LLMs. Our findings suggest that analyzing the topological structure of attention matrices can serve as an efficient and robust indicator of factual reliability in LLMs.
Large Language Models (LLMs) have revolutionized code generation, achieving exceptional results on various established benchmarking frameworks. However, concerns about data contamination - where benchmark data inadvertently leaks into pre-training or fine-tuning datasets - raise questions about the validity of these evaluations. While this issue is known, limiting the industrial adoption of LLM-driven software engineering, hardware coding has received little to no attention regarding these risks. For the first time, we analyze state-of-the-art (SOTA) evaluation frameworks for Verilog code generation (VerilogEval and RTLLM), using established methods for contamination detection (CCD and Min-K% Prob). We cover SOTA commercial and open-source LLMs (CodeGen2.5, Minitron 4b, Mistral 7b, phi-4 mini, LLaMA-{1,2,3.1}, GPT-{2,3.5,4o}, Deepseek-Coder, and CodeQwen 1.5), in baseline and fine-tuned models (RTLCoder and Verigen). Our study confirms that data contamination is a critical concern. We explore mitigations and the resulting trade-offs for code quality vs fairness (i.e., reducing contamination toward unbiased benchmarking).
Large Language Models (LLMs) are increasingly integrated into various daily tasks in Software Engineering such as coding and requirement elicitation. Despite their various capabilities and constant use, some interactions can lead to unexpected challenges (e.g. hallucinations or verbose answers) and, in turn, cause emotions that develop into frustration. Frustration can negatively impact engineers' productivity and well-being if they escalate into stress and burnout. In this paper, we assess the impact of LLM interactions on software engineers' emotional responses, specifically strains, and identify common causes of frustration when interacting with LLMs at work. Based on 62 survey responses from software engineers in industry and academia across various companies and universities, we found that a majority of our respondents experience frustrations or other related emotions regardless of the nature of their work. Additionally, our results showed that frustration mainly stemmed from issues with correctness and less critical issues such as adaptability to context or specific format. While such issues may not cause frustration in general, artefacts that do not follow certain preferences, standards, or best practices can make the output unusable without extensive modification, causing frustration over time. In addition to the frustration triggers, our study offers guidelines to improve the software engineers' experience, aiming to minimise long-term consequences on mental health.
Large Language Models (LLMs) are increasingly used as autonomous agents for multi-step tasks. However, most existing frameworks fail to maintain a structured understanding of the task state, often relying on linear prompt concatenation or shallow memory buffers. This leads to brittle performance, frequent hallucinations, and poor long-range coherence. In this work, we propose the Task Memory Engine (TME), a lightweight and structured memory module that tracks task execution using a hierarchical Task Memory Tree (TMT). Each node in the tree corresponds to a task step, storing relevant input, output, status, and sub-task relationships. We introduce a prompt synthesis method that dynamically generates LLM prompts based on the active node path, significantly improving execution consistency and contextual grounding. Through case studies and comparative experiments on multi-step agent tasks, we demonstrate that TME leads to better task completion accuracy and more interpretable behavior with minimal implementation overhead. A reference implementation of the core TME components is available at https://github.com/biubiutomato/TME-Agent, including basic examples and structured memory integration. While the current implementation uses a tree-based structure, TME is designed to be graph-aware, supporting reusable substeps, converging task paths, and shared dependencies. This lays the groundwork for future DAG-based memory architectures.
Large Language Models (LLMs) have shown impressive reasoning capabilities in well-defined problems with clear solutions, such as mathematics and coding. However, they still struggle with complex real-world scenarios like business negotiations, which require strategic reasoning-an ability to navigate dynamic environments and align long-term goals amidst uncertainty. Existing methods for strategic reasoning face challenges in adaptability, scalability, and transferring strategies to new contexts. To address these issues, we propose explicit policy optimization (EPO) for strategic reasoning, featuring an LLM that provides strategies in open-ended action space and can be plugged into arbitrary LLM agents to motivate goal-directed behavior. To improve adaptability and policy transferability, we train the strategic reasoning model via multi-turn reinforcement learning (RL) using process rewards and iterative self-play, without supervised fine-tuning (SFT) as a preliminary step. Experiments across social and physical domains demonstrate EPO's ability of long-term goal alignment through enhanced strategic reasoning, achieving state-of-the-art performance on social dialogue and web navigation tasks. Our findings reveal various collaborative reasoning mechanisms emergent in EPO and its effectiveness in generating novel strategies, underscoring its potential for strategic reasoning in real-world applications.
The training of large language models (LLMs) requires substantial computational resources, complex software stacks, and carefully designed workflows to achieve scalability and efficiency. This report presents best practices and insights gained from the OpenGPT-X project, a German initiative focused on developing open, multilingual LLMs optimized for European languages. We detail the use of high-performance computing (HPC) systems, primarily JUWELS Booster at JSC, for training Teuken-7B, a 7-billion-parameter transformer model. The report covers system architecture, training infrastructure, software choices, profiling and benchmarking tools, as well as engineering and operational challenges.
Studies on evaluation metrics and LLM-as-a-Judge models for automatic text summarization have largely been focused on English, limiting our understanding of their effectiveness in other languages. Through our new dataset BASSE (BAsque and Spanish Summarization Evaluation), we address this situation by collecting human judgments on 2,040 abstractive summaries in Basque and Spanish, generated either manually or by five LLMs with four different prompts. For each summary, annotators evaluated five criteria on a 5-point Likert scale: coherence, consistency, fluency, relevance, and 5W1H. We use these data to reevaluate traditional automatic metrics used for evaluating summaries, as well as several LLM-as-a-Judge models that show strong performance on this task in English. Our results show that currently proprietary judge LLMs have the highest correlation with human judgments, followed by criteria-specific automatic metrics, while open-sourced judge LLMs perform poorly. We release BASSE and our code publicly, along with the first large-scale Basque summarization dataset containing 22,525 news articles with their subheads.
As Large Language Models (LLMs) become integral to scientific workflows, concerns over the confidentiality and ethical handling of confidential data have emerged. This paper explores data exposure risks through LLM-powered scientific tools, which can inadvertently leak confidential information, including intellectual property and proprietary data, from scientists' perspectives. We propose "DataShield", a framework designed to detect confidential data leaks, summarize privacy policies, and visualize data flow, ensuring alignment with organizational policies and procedures. Our approach aims to inform scientists about data handling practices, enabling them to make informed decisions and protect sensitive information. Ongoing user studies with scientists are underway to evaluate the framework's usability, trustworthiness, and effectiveness in tackling real-world privacy challenges.
Efficient inference of large language models (LLMs) is hindered by an ever-growing key-value (KV) cache, making KV cache compression a critical research direction. Traditional methods selectively evict less important KV cache entries based on attention scores or position heuristics, which leads to information loss and hallucinations. Recently, merging-based strategies have been explored to retain more information by merging KV pairs that would be discarded; however, these existing approaches inevitably introduce inconsistencies in attention distributions before and after merging, causing output perturbation and degraded generation quality. To overcome this challenge, we propose KeepKV, a novel adaptive KV cache merging method designed to eliminate output perturbation while preserving performance under strict memory constraints. KeepKV introduces the Electoral Votes mechanism that records merging history and adaptively adjusts attention scores. Moreover, it further leverages a novel Zero Inference-Perturbation Merging methods, keeping attention consistency and compensating for attention loss resulting from cache merging. KeepKV successfully retains essential context information within a significantly compressed cache. Extensive experiments on various benchmarks and LLM architectures demonstrate that KeepKV substantially reduces memory usage, enhances inference throughput by more than 2x and keeps superior generation quality even with 10% KV cache budgets.
Being able to use tools is a widely recognised indicator of intelligence across species. Humans, for instance, have demonstrated mastery of tool use for over two million years. The ability to use tools is invaluable as it extends an organism's reach and enhances its capacity to interact with objects and the environment. Being able to understand the geometric-mechanical relations between the tools-objects-environments allows certain species (e.g., apes and crows) to reach food in narrow constrained spaces. The same principles of physical augmentation and its associated non-prehensile manipulation capabilities also apply to robotic systems. For example, by instrumenting them with different types of end-effectors, robots can (in principle) dexterously interact (e.g., push and flip) with objects of various shapes and masses akin to its biological counterpart. However, developing this type of manipulation skill is still an open research problem. Furthermore, the complexity of planning tool-object manipulation tasks, particularly in coordinating the actions of dual-arm robots, presents significant challenges. To address these complexities, we propose integrating Large Language Models (LLMs) to assist in planning and executing these intricate manipulations, thereby enhancing the robot's ability to perform in diverse scenarios.
The limited reasoning capabilities of small language models (SLMs) cast doubt on their suitability for tasks demanding deep, multi-step logical deduction. This paper introduces a framework called Small Reasons, Large Hints (SMART), which selectively augments SLM reasoning with targeted guidance from large language models (LLMs). Inspired by the concept of cognitive scaffolding, SMART employs a score-based evaluation to identify uncertain reasoning steps and injects corrective LLM-generated reasoning only when necessary. By framing structured reasoning as an optimal policy search, our approach steers the reasoning trajectory toward correct solutions without exhaustive sampling. Our experiments on mathematical reasoning datasets demonstrate that targeted external scaffolding significantly improves performance, paving the way for collaborative use of both SLM and LLM to tackle complex reasoning tasks that are currently unsolvable by SLMs alone.
Effective pest management is complex due to the need for accurate, context-specific decisions. Recent advancements in large language models (LLMs) open new possibilities for addressing these challenges by providing sophisticated, adaptive knowledge acquisition and reasoning. However, existing LLM-based pest management approaches often rely on a single-agent paradigm, which can limit their capacity to incorporate diverse external information, engage in systematic validation, and address complex, threshold-driven decisions. To overcome these limitations, we introduce PestMA, an LLM-based multi-agent system (MAS) designed to generate reliable and evidence-based pest management advice. Building on an editorial paradigm, PestMA features three specialized agents, an Editor for synthesizing pest management recommendations, a Retriever for gathering relevant external data, and a Validator for ensuring correctness. Evaluations on real-world pest scenarios demonstrate that PestMA achieves an initial accuracy of 86.8% for pest management decisions, which increases to 92.6% after validation. These results underscore the value of collaborative agent-based workflows in refining and validating decisions, highlighting the potential of LLM-based multi-agent systems to automate and enhance pest management processes.
Building high-quality datasets and labeling query-document relevance are essential yet resource-intensive tasks, requiring detailed guidelines and substantial effort from human annotators. This paper explores the use of small, fine-tuned large language models (LLMs) to automate relevance assessment, with a focus on improving ranking models' performance by augmenting their training dataset. We fine-tuned small LLMs to enhance relevance assessments, thereby improving dataset creation quality for downstream ranking model training. Our experiments demonstrate that these fine-tuned small LLMs not only outperform certain closed source models on our dataset but also lead to substantial improvements in ranking model performance. These results highlight the potential of leveraging small LLMs for efficient and scalable dataset augmentation, providing a practical solution for search engine optimization.
The reasoning capabilities of large language models (LLMs), such as OpenAI's o1 and DeepSeek-R1, have seen substantial advancements through deep thinking. However, these enhancements come with significant resource demands, underscoring the need to explore strategies to train effective reasoning LLMs with far fewer parameters. A critical challenge is that smaller models have different capacities and cognitive trajectories than their larger counterparts. Hence, direct distillation of chain-of-thought (CoT) results from large LLMs to smaller ones can be sometimes ineffective and requires a huge amount of annotated data. In this paper, we introduce a novel framework called Critique-Rethink-Verify (CRV), designed for training smaller yet powerful reasoning LLMs. Our CRV framework consists of multiple LLM agents, each specializing in unique abilities: (i) critiquing the CoTs according to the cognitive capabilities of smaller models, (ii) rethinking and refining these CoTs based on the critiques, and (iii) verifying the correctness of the refined results. We further propose the cognitive preference optimization (CogPO) algorithm to enhance the reasoning abilities of smaller models by aligning thoughts of these models with their cognitive capacities. Comprehensive evaluations on challenging reasoning benchmarks demonstrate the efficacy of CRV and CogPO, which outperforms other training methods by a large margin.
Large Language Models (LLMs) often struggle with code generation tasks involving niche software libraries. Existing code generation techniques with only human-oriented documentation can fail -- even when the LLM has access to web search and the library is documented online. To address this challenge, we propose ReadMe.LLM, LLM-oriented documentation for software libraries. By attaching the contents of ReadMe.LLM to a query, performance consistently improves to near-perfect accuracy, with one case study demonstrating up to 100% success across all tested models. We propose a software development lifecycle where LLM-specific documentation is maintained alongside traditional software updates. In this study, we present two practical applications of the ReadMe.LLM idea with diverse software libraries, highlighting that our proposed approach could generalize across programming domains.
As the prevalence of mental health crises increases on social media platforms, identifying and preventing potential harm has become an urgent challenge. This study introduces a large language model (LLM)-based text transfer recognition method for social network crisis intervention, enhanced with domain-specific mental health knowledge. We propose a multi-level framework that incorporates transfer learning using BERT, and integrates mental health knowledge, sentiment analysis, and behavior prediction techniques. The framework includes a crisis annotation tool trained on social media datasets from real-world events, enabling the model to detect nuanced emotional cues and identify psychological crises. Experimental results show that the proposed method outperforms traditional models in crisis detection accuracy and exhibits greater sensitivity to subtle emotional and contextual variations.
Deploying large language models (LLMs) presents critical challenges due to the inherent trade-offs associated with key performance metrics, such as latency, accuracy, and throughput. Typically, gains in one metric is accompanied with degradation in others. Early-Exit LLMs (EE-LLMs) efficiently navigate this trade-off space by skipping some of the later model layers when it confidently finds an output token early, thus reducing latency without impacting accuracy. However, as the early exits taken depend on the task and are unknown apriori to request processing, EE-LLMs conservatively load the entire model, limiting resource savings and throughput. Also, current frameworks statically select a model for a user task, limiting our ability to adapt to changing nature of the input queries. We propose HELIOS to address these challenges. First, HELIOS shortlists a set of candidate LLMs, evaluates them using a subset of prompts, gathering telemetry data in real-time. Second, HELIOS uses the early exit data from these evaluations to greedily load the selected model only up to a limited number of layers. This approach yields memory savings which enables us to process more requests at the same time, thereby improving throughput. Third, HELIOS monitors and periodically reassesses the performance of the candidate LLMs and if needed, switches to another model that can service incoming queries more efficiently (such as using fewer layers without lowering accuracy). Our evaluations show that HELIOS achieves 1.48$\times$ throughput, 1.10$\times$ energy-efficiency, 1.39$\times$ lower response time, and 3.7$\times$ improvements in inference batch sizes compared to the baseline, when optimizing for the respective service level objectives.
Common Vulnerability and Exposure (CVE) records are fundamental to cybersecurity, offering unique identifiers for publicly known software and system vulnerabilities. Each CVE is typically assigned a Common Vulnerability Scoring System (CVSS) score to support risk prioritization and remediation. However, score inconsistencies often arise due to subjective interpretations of certain metrics. As the number of new CVEs continues to grow rapidly, automation is increasingly necessary to ensure timely and consistent scoring. While prior studies have explored automated methods, the application of Large Language Models (LLMs), despite their recent popularity, remains relatively underexplored. In this work, we evaluate the effectiveness of LLMs in generating CVSS scores for newly reported vulnerabilities. We investigate various prompt engineering strategies to enhance their accuracy and compare LLM-generated scores against those from embedding-based models, which use vector representations classified via supervised learning. Our results show that while LLMs demonstrate potential in automating CVSS evaluation, embedding-based methods outperform them in scoring more subjective components, particularly confidentiality, integrity, and availability impacts. These findings underscore the complexity of CVSS scoring and suggest that combining LLMs with embedding-based methods could yield more reliable results across all scoring components.
This paper introduces GestureCoach, a system designed to help speakers deliver more engaging talks by guiding them to gesture effectively during rehearsal. GestureCoach combines an LLM-driven gesture recommendation model with a rehearsal interface that proactively cues speakers to gesture appropriately. Trained on experts' gesturing patterns from TED talks, the model consists of two modules: an emphasis proposal module, which predicts when to gesture by identifying gesture-worthy text segments in the presenter notes, and a gesture identification module, which determines what gesture to use by retrieving semantically appropriate gestures from a curated gesture database. Results of a model performance evaluation and user study (N=30) show that the emphasis proposal module outperforms off-the-shelf LLMs in identifying suitable gesture regions, and that participants rated the majority of these predicted regions and their corresponding gestures as highly appropriate. A subsequent user study (N=10) showed that rehearsing with GestureCoach encouraged speakers to gesture and significantly increased gesture diversity, resulting in more engaging talks. We conclude with design implications for future AI-driven rehearsal systems.
While recent advances in large language models (LLMs) have significantly enhanced performance across diverse natural language tasks, the high computational and financial costs associated with their deployment remain substantial barriers. Existing routing strategies partially alleviate this challenge by assigning queries to cheaper or specialized models, but they frequently rely on extensive labeled data or fragile task-specific heuristics. Conversely, fusion techniques aggregate multiple LLM outputs to boost accuracy and robustness, yet they often exacerbate cost and may reinforce shared biases. We introduce EMAFusion, a new framework that self-optimizes for seamless LLM selection and reliable execution for a given query. Specifically, EMAFusion integrates a taxonomy-based router for familiar query types, a learned router for ambiguous inputs, and a cascading approach that progressively escalates from cheaper to more expensive models based on multi-judge confidence evaluations. Through extensive evaluations, we find EMAFusion outperforms the best individual models by over 2.6 percentage points (94.3% vs. 91.7%), while being 4X cheaper than the average cost. EMAFusion further achieves a remarkable 17.1 percentage point improvement over models like GPT-4 at less than 1/20th the cost. Our combined routing approach delivers 94.3% accuracy compared to taxonomy-based (88.1%) and learned model predictor-based (91.7%) methods alone, demonstrating the effectiveness of our unified strategy. Finally, EMAFusion supports flexible cost-accuracy trade-offs, allowing users to balance their budgetary constraints and performance needs.
This paper introduces an LLM-based Latin-to-English translation platform designed to address the challenges of translating Latin texts. We named the model LITERA, which stands for Latin Interpretation and Translations into English for Research Assistance. Through a multi-layered translation process utilizing a fine-tuned version of GPT-4o-mini and GPT-4o, LITERA offers an unprecedented level of accuracy, showcased by greatly improved BLEU scores, particularly in classical Latin, along with improved BLEURT scores. The development of LITERA involved close collaboration with Duke University's Classical Studies Department, which was instrumental in creating a small, high-quality parallel Latin-English dataset. This paper details the architecture, fine-tuning methodology, and prompting strategies used in LITERA, emphasizing its ability to produce literal translations.
Large language models (LLMs) have demonstrated remarkable reasoning capabilities when prompted with strategies such as Chain-of-Thought (CoT). However, these approaches focus on token-level output without considering internal weight dynamics. We introduce Weight-of-Thought (WoT) reasoning, a novel approach that examines neural network weights before inference to identify reasoning pathways. Unlike existing methods, WoT explores the weight space through graph-based message passing, multi-step reasoning processes, and attention mechanisms. Our implementation creates an interconnected graph of reasoning nodes. Experiments on diverse reasoning tasks (syllogistic, mathematical, algebraic, combinatorial, and geometric) demonstrate that WoT achieves superior performance compared to traditional methods, particularly for complex problems. This approach leads to both improved performance and greater interpretability of the reasoning process, offering a promising direction for enhancing LLM reasoning capabilities.
Understanding code represents a core ability needed for automating software development tasks. While foundation models like LLMs show impressive results across many software engineering challenges, the extent of their true semantic understanding beyond simple token recognition remains unclear. This research uses code obfuscation as a structured testing framework to evaluate LLMs' semantic understanding capabilities. We methodically apply controlled obfuscation changes to source code and measure comprehension through two complementary tasks: generating accurate descriptions of obfuscated code and performing deobfuscation, a skill with important implications for reverse engineering applications. Our testing approach includes 13 cutting-edge models, covering both code-specialized (e.g., StarCoder2) and general-purpose (e.g., GPT-4o) architectures, evaluated on a benchmark created from CodeNet and consisting of filtered 250 Java programming problems and their solutions. Findings show a statistically significant performance decline as obfuscation complexity increases, with unexpected resilience shown by general-purpose models compared to their code-focused counterparts. While some models successfully identify obfuscation techniques, their ability to reconstruct the underlying program logic remains constrained, suggesting limitations in their semantic representation mechanisms. This research introduces a new evaluation approach for assessing code comprehension in language models and establishes empirical baselines for advancing research in security-critical code analysis applications such as reverse engineering and adversarial code analysis.
The design of Analog and Mixed-Signal (AMS) integrated circuits (ICs) often involves significant manual effort, especially during the transistor sizing process. While Machine Learning techniques in Electronic Design Automation (EDA) have shown promise in reducing complexity and minimizing human intervention, they still face challenges such as numerous iterations and a lack of knowledge about AMS circuit design. Recently, Large Language Models (LLMs) have demonstrated significant potential across various fields, showing a certain level of knowledge in circuit design and indicating their potential to automate the transistor sizing process. In this work, we propose an LLM-based AI agent for AMS circuit design to assist in the sizing process. By integrating LLMs with external circuit simulation tools and data analysis functions and employing prompt engineering strategies, the agent successfully optimized multiple circuits to achieve target performance metrics. We evaluated the performance of different LLMs to assess their applicability and optimization effectiveness across seven basic circuits, and selected the best-performing model Claude 3.5 Sonnet for further exploration on an operational amplifier, with complementary input stage and class AB output stage. This circuit was evaluated against nine performance metrics, and we conducted experiments under three distinct performance requirement groups. A success rate of up to 60% was achieved for reaching the target requirements. Overall, this work demonstrates the potential of LLMs to improve AMS circuit design.
The emergence of generative AI chatbots such as ChatGPT has prompted growing public and academic interest in their role as informal mental health support tools. While early rule-based systems have been around for several years, large language models (LLMs) offer new capabilities in conversational fluency, empathy simulation, and availability. This study explores how users engage with LLMs as mental health tools by analyzing over 10,000 TikTok comments from videos referencing LLMs as mental health tools. Using a self-developed tiered coding schema and supervised classification models, we identify user experiences, attitudes, and recurring themes. Results show that nearly 20% of comments reflect personal use, with these users expressing overwhelmingly positive attitudes. Commonly cited benefits include accessibility, emotional support, and perceived therapeutic value. However, concerns around privacy, generic responses, and the lack of professional oversight remain prominent. It is important to note that the user feedback does not indicate which therapeutic framework, if any, the LLM-generated output aligns with. While the findings underscore the growing relevance of AI in everyday practices, they also highlight the urgent need for clinical and ethical scrutiny in the use of AI for mental health support.
WebShell attacks, in which malicious scripts are injected into web servers, are a major cybersecurity threat. Traditional machine learning and deep learning methods are hampered by issues such as the need for extensive training data, catastrophic forgetting, and poor generalization. Recently, Large Language Models (LLMs) have gained attention for code-related tasks, but their potential in WebShell detection remains underexplored. In this paper, we make two major contributions: (1) a comprehensive evaluation of seven LLMs, including GPT-4, LLaMA 3.1 70B, and Qwen 2.5 variants, benchmarked against traditional sequence- and graph-based methods using a dataset of 26.59K PHP scripts, and (2) the Behavioral Function-Aware Detection (BFAD) framework, designed to address the specific challenges of applying LLMs to this domain. Our framework integrates three components: a Critical Function Filter that isolates malicious PHP function calls, a Context-Aware Code Extraction strategy that captures the most behaviorally indicative code segments, and Weighted Behavioral Function Profiling (WBFP) that enhances in-context learning by prioritizing the most relevant demonstrations based on discriminative function-level profiles. Our results show that larger LLMs achieve near-perfect precision but lower recall, while smaller models exhibit the opposite trade-off. However, all models lag behind previous State-Of-The-Art (SOTA) methods. With BFAD, the performance of all LLMs improved, with an average F1 score increase of 13.82%. Larger models such as GPT-4, LLaMA 3.1 70B, and Qwen 2.5 14B outperform SOTA methods, while smaller models such as Qwen 2.5 3B achieve performance competitive with traditional methods. This work is the first to explore the feasibility and limitations of LLMs for WebShell detection, and provides solutions to address the challenges in this task.
We present a system using Multimodal LLMs (MLLMs) to analyze a large database with tens of millions of images captured at different times, with the aim of discovering patterns in temporal changes. Specifically, we aim to capture frequent co-occurring changes ("trends") across a city over a certain period. Unlike previous visual analyses, our analysis answers open-ended queries (e.g., "what are the frequent types of changes in the city?") without any predetermined target subjects or training labels. These properties cast prior learning-based or unsupervised visual analysis tools unsuitable. We identify MLLMs as a novel tool for their open-ended semantic understanding capabilities. Yet, our datasets are four orders of magnitude too large for an MLLM to ingest as context. So we introduce a bottom-up procedure that decomposes the massive visual analysis problem into more tractable sub-problems. We carefully design MLLM-based solutions to each sub-problem. During experiments and ablation studies with our system, we find it significantly outperforms baselines and is able to discover interesting trends from images captured in large cities (e.g., "addition of outdoor dining,", "overpass was painted blue," etc.). See more results and interactive demos at https://boyangdeng.com/visual-chronicles.
Recent advancements in Large Language Models (LLMs) have enabled them to approach human-level persuasion capabilities. However, such potential also raises concerns about the safety risks of LLM-driven persuasion, particularly their potential for unethical influence through manipulation, deception, exploitation of vulnerabilities, and many other harmful tactics. In this work, we present a systematic investigation of LLM persuasion safety through two critical aspects: (1) whether LLMs appropriately reject unethical persuasion tasks and avoid unethical strategies during execution, including cases where the initial persuasion goal appears ethically neutral, and (2) how influencing factors like personality traits and external pressures affect their behavior. To this end, we introduce PersuSafety, the first comprehensive framework for the assessment of persuasion safety which consists of three stages, i.e., persuasion scene creation, persuasive conversation simulation, and persuasion safety assessment. PersuSafety covers 6 diverse unethical persuasion topics and 15 common unethical strategies. Through extensive experiments across 8 widely used LLMs, we observe significant safety concerns in most LLMs, including failing to identify harmful persuasion tasks and leveraging various unethical persuasion strategies. Our study calls for more attention to improve safety alignment in progressive and goal-driven conversations such as persuasion.
Knowledge editing has emerged as an effective approach for updating large language models (LLMs) by modifying their internal knowledge. However, their application to the biomedical domain faces unique challenges due to the long-tailed distribution of biomedical knowledge, where rare and infrequent information is prevalent. In this paper, we conduct the first comprehensive study to investigate the effectiveness of knowledge editing methods for editing long-tail biomedical knowledge. Our results indicate that, while existing editing methods can enhance LLMs' performance on long-tail biomedical knowledge, their performance on long-tail knowledge remains inferior to that on high-frequency popular knowledge, even after editing. Our further analysis reveals that long-tail biomedical knowledge contains a significant amount of one-to-many knowledge, where one subject and relation link to multiple objects. This high prevalence of one-to-many knowledge limits the effectiveness of knowledge editing in improving LLMs' understanding of long-tail biomedical knowledge, highlighting the need for tailored strategies to bridge this performance gap.
Objective: This study investigates the potential of Large Language Models (LLMs) as an alternative to human expert elicitation for extracting structured causal knowledge and facilitating causal modeling in biometric and healthcare applications. Material and Methods: LLM-generated causal structures, specifically Bayesian networks (BNs), were benchmarked against traditional statistical methods (e.g., Bayesian Information Criterion) using healthcare datasets. Validation techniques included structural equation modeling (SEM) to verifying relationships, and measures such as entropy, predictive accuracy, and robustness to compare network structures. Results and Discussion: LLM-generated BNs demonstrated lower entropy than expert-elicited and statistically generated BNs, suggesting higher confidence and precision in predictions. However, limitations such as contextual constraints, hallucinated dependencies, and potential biases inherited from training data require further investigation. Conclusion: LLMs represent a novel frontier in expert elicitation for probabilistic causal modeling, promising to improve transparency and reduce uncertainty in the decision-making using such models.
Crafting a marketing message (copy), or copywriting is a challenging generation task, as the copy must adhere to various constraints. Copy creation is inherently iterative for humans, starting with an initial draft followed by successive refinements. However, manual copy creation is time-consuming and expensive, resulting in only a few copies for each use case. This limitation restricts our ability to personalize content to customers. Contrary to the manual approach, LLMs can generate copies quickly, but the generated content does not consistently meet all the constraints on the first attempt (similar to humans). While recent studies have shown promise in improving constrained generation through iterative refinement, they have primarily addressed tasks with only a few simple constraints. Consequently, the effectiveness of iterative refinement for tasks such as copy generation, which involves many intricate constraints, remains unclear. To address this gap, we propose an LLM-based end-to-end framework for scalable copy generation using iterative refinement. To the best of our knowledge, this is the first study to address multiple challenging constraints simultaneously in copy generation. Examples of these constraints include length, topics, keywords, preferred lexical ordering, and tone of voice. We demonstrate the performance of our framework by creating copies for e-commerce banners for three different use cases of varying complexity. Our results show that iterative refinement increases the copy success rate by $16.25-35.91$% across use cases. Furthermore, the copies generated using our approach outperformed manually created content in multiple pilot studies using a multi-armed bandit framework. The winning copy improved the click-through rate by $38.5-45.21$%.
Optimizing Register Transfer Level (RTL) code is crucial for improving the power, performance, and area (PPA) of digital circuits in the early stages of synthesis. Manual rewriting, guided by synthesis feedback, can yield high-quality results but is time-consuming and error-prone. Most existing compiler-based approaches have difficulty handling complex design constraints. Large Language Model (LLM)-based methods have emerged as a promising alternative to address these challenges. However, LLM-based approaches often face difficulties in ensuring alignment between the generated code and the provided prompts. This paper presents SymRTLO, a novel neuron-symbolic RTL optimization framework that seamlessly integrates LLM-based code rewriting with symbolic reasoning techniques. Our method incorporates a retrieval-augmented generation (RAG) system of optimization rules and Abstract Syntax Tree (AST)-based templates, enabling LLM-based rewriting that maintains syntactic correctness while minimizing undesired circuit behaviors. A symbolic module is proposed for analyzing and optimizing finite state machine (FSM) logic, allowing fine-grained state merging and partial specification handling beyond the scope of pattern-based compilers. Furthermore, a fast verification pipeline, combining formal equivalence checks with test-driven validation, further reduces the complexity of verification. Experiments on the RTL-Rewriter benchmark with Synopsys Design Compiler and Yosys show that SymRTLO improves power, performance, and area (PPA) by up to 43.9%, 62.5%, and 51.1%, respectively, compared to the state-of-the-art methods.
We present MultiLoKo, a new benchmark for evaluating multilinguality in LLMs covering 31 languages. MultiLoKo consists of three partitions: a main partition consisting of 500 questions per language, separately sourced to be locally relevant to the specific language, and two translated partitions, containing human-authored translations from 30 non-English languages to English and vice versa. For comparison, we also release corresponding machine-authored translations. The data is equally distributed over two splits: a dev split and a blind, out-of-distribution test split. MultiLoKo can be used to study a variety of questions regarding the multilinguality of LLMs as well as meta-questions about multilingual benchmark creation. We compute MultiLoKo scores for 11 base and chat models marketed to be multilingual and study their average performance, their performance parity across languages, how much their ability to answer questions depends on the question language, and which languages are most difficult. None of the models we studied performs well on MultiLoKo, as indicated by low average scores as well as large differences between the best and worst scoring languages. Furthermore, we find a substantial effect of the question language, indicating sub-optimal knowledge transfer between languages. Lastly, we find that using local vs English-translated data can result in differences more than 20 points for the best performing models, drastically change the estimated difficulty of some languages. For using machines instead of human translations, we find a weaker effect on ordering of language difficulty, a larger difference in model rankings, and a substantial drop in estimated performance for all models.
AlayaDB is a cutting-edge vector database system natively architected for efficient and effective long-context inference for Large Language Models (LLMs) at AlayaDB AI. Specifically, it decouples the KV cache and attention computation from the LLM inference systems, and encapsulates them into a novel vector database system. For the Model as a Service providers (MaaS), AlayaDB consumes fewer hardware resources and offers higher generation quality for various workloads with different kinds of Service Level Objectives (SLOs), when comparing with the existing alternative solutions (e.g., KV cache disaggregation, retrieval-based sparse attention). The crux of AlayaDB is that it abstracts the attention computation and cache management for LLM inference into a query processing procedure, and optimizes the performance via a native query optimizer. In this work, we demonstrate the effectiveness of AlayaDB via (i) three use cases from our industry partners, and (ii) extensive experimental results on LLM inference benchmarks.
Large language models (LLMs) have shown impressive potential in helping with numerous medical challenges. Deploying LLMs in high-stakes applications such as medicine, however, brings in many concerns. One major area of concern relates to biased behaviors of LLMs in medical applications, leading to unfair treatment of individuals. To pave the way for the responsible and impactful deployment of Med LLMs, rigorous evaluation is a key prerequisite. Due to the huge complexity and variability of different medical scenarios, existing work in this domain has primarily relied on using manually crafted datasets for bias evaluation. In this study, we present a new method to scale up such bias evaluations by automatically generating test cases based on rigorous medical evidence. We specifically target the challenges of a) domain-specificity of bias characterization, b) hallucinating while generating the test cases, and c) various dependencies between the health outcomes and sensitive attributes. To that end, we offer new methods to address these challenges integrated with our generative pipeline, using medical knowledge graphs, medical ontologies, and customized general LLM evaluation frameworks in our method. Through a series of extensive experiments, we show that the test cases generated by our proposed method can effectively reveal bias patterns in Med LLMs at larger and more flexible scales than human-crafted datasets. We publish a large bias evaluation dataset using our pipeline, which is dedicated to a few medical case studies. A live demo of our application for vignette generation is available at https://vignette.streamlit.app. Our code is also available at https://github.com/healthylaife/autofair.
Current safety alignment techniques for large language models (LLMs) face two key challenges: (1) under-generalization, which leaves models vulnerable to novel jailbreak attacks, and (2) over-alignment, which leads to the excessive refusal of benign instructions. Our preliminary investigation reveals semantic overlap between jailbreak/harmful queries and normal prompts in embedding space, suggesting that more effective safety alignment requires a deeper semantic understanding. This motivates us to incorporate safety-policy-driven reasoning into the alignment process. To this end, we propose the Safety-oriented Reasoning Optimization Framework (SaRO), which consists of two stages: (1) Reasoning-style Warmup (RW) that enables LLMs to internalize long-chain reasoning through supervised fine-tuning, and (2) Safety-oriented Reasoning Process Optimization (SRPO) that promotes safety reflection via direct preference optimization (DPO). Extensive experiments demonstrate the superiority of SaRO over traditional alignment methods.
Usability testing is a fundamental research method that user experience (UX) researchers use to evaluate and iterate a web design, but\textbf{ how to evaluate and iterate the usability testing study design } itself? Recent advances in Large Language Model-simulated Agent (\textbf{LLM Agent}) research inspired us to design \textbf{UXAgent} to support UX researchers in evaluating and reiterating their usability testing study design before they conduct the real human-subject study. Our system features a Persona Generator module, an LLM Agent module, and a Universal Browser Connector module to automatically generate thousands of simulated users to interactively test the target website. The system also provides an Agent Interview Interface and a Video Replay Interface so that the UX researchers can easily review and analyze the generated qualitative and quantitative log data. Through a heuristic evaluation, five UX researcher participants praised the innovation of our system but also expressed concerns about the future of LLM Agent usage in UX studies.
Large language models (LLMs) pre-trained predominantly on English text exhibit surprising multilingual capabilities, yet the mechanisms driving cross-lingual generalization remain poorly understood. This work investigates how the alignment of representations for text written in different languages correlates with LLM performance on natural language understanding tasks and translation tasks, both at the language and the instance level. For this purpose, we introduce cross-lingual alignment metrics such as the Discriminative Alignment Index (DALI) to quantify the alignment at an instance level for discriminative tasks. Through experiments on three natural language understanding tasks (Belebele, XStoryCloze, XCOPA), and machine translation, we find that while cross-lingual alignment metrics strongly correlate with task accuracy at the language level, the sample-level alignment often fails to distinguish correct from incorrect predictions, exposing alignment as a necessary but insufficient condition for success.
The burgeoning presence of Large Language Models (LLM) is propelling the development of personalized recommender systems. Most existing LLM-based methods fail to sufficiently explore the multi-view graph structure correlations inherent in recommendation scenarios. To this end, we propose a novel framework, Hypergraph Enhanced LLM Learning for multimodal Recommendation (HeLLM), designed to equip LLMs with the capability to capture intricate higher-order semantic correlations by fusing graph-level contextual signals with sequence-level behavioral patterns. In the recommender pre-training phase, we design a user hypergraph to uncover shared interest preferences among users and an item hypergraph to capture correlations within multimodal similarities among items. The hypergraph convolution and synergistic contrastive learning mechanism are introduced to enhance the distinguishability of learned representations. In the LLM fine-tuning phase, we inject the learned graph-structured embeddings directly into the LLM's architecture and integrate sequential features capturing each user's chronological behavior. This process enables hypergraphs to leverage graph-structured information as global context, enhancing the LLM's ability to perceive complex relational patterns and integrate multimodal information, while also modeling local temporal dynamics. Extensive experiments demonstrate the superiority of our proposed method over state-of-the-art baselines, confirming the advantages of fusing hypergraph-based context with sequential user behavior in LLMs for recommendation.
The evaluation of open-ended responses in serious games presents a unique challenge, as correctness is often subjective. Large Language Models (LLMs) are increasingly being explored as evaluators in such contexts, yet their accuracy and consistency remain uncertain, particularly for smaller models intended for local execution. This study investigates the reliability of five small-scale LLMs when assessing player responses in \textit{En-join}, a game that simulates decision-making within energy communities. By leveraging traditional binary classification metrics (including accuracy, true positive rate, and true negative rate), we systematically compare these models across different evaluation scenarios. Our results highlight the strengths and limitations of each model, revealing trade-offs between sensitivity, specificity, and overall performance. We demonstrate that while some models excel at identifying correct responses, others struggle with false positives or inconsistent evaluations. The findings highlight the need for context-aware evaluation frameworks and careful model selection when deploying LLMs as evaluators. This work contributes to the broader discourse on the trustworthiness of AI-driven assessment tools, offering insights into how different LLM architectures handle subjective evaluation tasks.
Span-level emotion-cause-category triplet extraction represents a novel and complex challenge within emotion cause analysis. This task involves identifying emotion spans, cause spans, and their associated emotion categories within the text to form structured triplets. While prior research has predominantly concentrated on clause-level emotion-cause pair extraction and span-level emotion-cause detection, these methods often confront challenges originating from redundant information retrieval and difficulty in accurately determining emotion categories, particularly when emotions are expressed implicitly or ambiguously. To overcome these challenges, this study explores a fine-grained approach to span-level emotion-cause-category triplet extraction and introduces an innovative framework that leverages instruction tuning and data augmentation techniques based on large language models. The proposed method employs task-specific triplet extraction instructions and utilizes low-rank adaptation to fine-tune large language models, eliminating the necessity for intricate task-specific architectures. Furthermore, a prompt-based data augmentation strategy is developed to address data scarcity by guiding large language models in generating high-quality synthetic training data. Extensive experimental evaluations demonstrate that the proposed approach significantly outperforms existing baseline methods, achieving at least a 12.8% improvement in span-level emotion-cause-category triplet extraction metrics. The results demonstrate the method's effectiveness and robustness, offering a promising avenue for advancing research in emotion cause analysis. The source code is available at https://github.com/zxgnlp/InstruDa-LLM.
Large language models (LLMs) have demonstrated revolutionary capabilities in understanding complex contexts and performing a wide range of tasks. However, LLMs can also answer questions that are unethical or harmful, raising concerns about their applications. To regulate LLMs' responses to such questions, a training strategy called \textit{alignment} can help. Yet, alignment can be unexpectedly compromised when fine-tuning an LLM for downstream tasks. This paper focuses on recovering the alignment lost during fine-tuning. We observe that there are two distinct directions inherent in an aligned LLM: the \textit{aligned direction} and the \textit{harmful direction}. An LLM is inclined to answer questions in the aligned direction while refusing queries in the harmful direction. Therefore, we propose to recover the harmful direction of the fine-tuned model that has been compromised. Specifically, we restore a small subset of the fine-tuned model's weight parameters from the original aligned model using gradient descent. We also introduce a rollback mechanism to avoid aggressive recovery and maintain downstream task performance. Our evaluation on 125 fine-tuned LLMs demonstrates that our method can reduce their harmful rate (percentage of answering harmful questions) from 33.25\% to 1.74\%, without sacrificing task performance much. In contrast, the existing methods either only reduce the harmful rate to a limited extent or significantly impact the normal functionality. Our code is available at https://github.com/kangyangWHU/LLMAlignment
Peer review at AI conferences is stressed by rapidly rising submission volumes, leading to deteriorating review quality and increased author dissatisfaction. To address these issues, we developed Review Feedback Agent, a system leveraging multiple large language models (LLMs) to improve review clarity and actionability by providing automated feedback on vague comments, content misunderstandings, and unprofessional remarks to reviewers. Implemented at ICLR 2025 as a large randomized control study, our system provided optional feedback to more than 20,000 randomly selected reviews. To ensure high-quality feedback for reviewers at this scale, we also developed a suite of automated reliability tests powered by LLMs that acted as guardrails to ensure feedback quality, with feedback only being sent to reviewers if it passed all the tests. The results show that 27% of reviewers who received feedback updated their reviews, and over 12,000 feedback suggestions from the agent were incorporated by those reviewers. This suggests that many reviewers found the AI-generated feedback sufficiently helpful to merit updating their reviews. Incorporating AI feedback led to significantly longer reviews (an average increase of 80 words among those who updated after receiving feedback) and more informative reviews, as evaluated by blinded researchers. Moreover, reviewers who were selected to receive AI feedback were also more engaged during paper rebuttals, as seen in longer author-reviewer discussions. This work demonstrates that carefully designed LLM-generated review feedback can enhance peer review quality by making reviews more specific and actionable while increasing engagement between reviewers and authors. The Review Feedback Agent is publicly available at https://github.com/zou-group/review_feedback_agent.
Test-time scaling has significantly improved large language model performance, enabling deeper reasoning to solve complex problems. However, this increased reasoning capability also leads to excessive token generation and unnecessary problem-solving attempts. We introduce Don\'t Answer Bench (DNA Bench), a new benchmark designed to evaluate LLMs ability to robustly understand the tricky reasoning triggers and avoiding unnecessary generation. DNA Bench consists of 150 adversarially designed prompts that are easy for humans to understand and respond to, but surprisingly not for many of the recent prominent LLMs. DNA Bench tests models abilities across different capabilities, such as instruction adherence, hallucination avoidance, redundancy filtering, and unanswerable question recognition. We evaluate reasoning LLMs (RLMs), including DeepSeek-R1, OpenAI O3-mini, Claude-3.7-sonnet and compare them against a powerful non-reasoning model, e.g., GPT-4o. Our experiments reveal that RLMs generate up to 70x more tokens than necessary, often failing at tasks that simpler non-reasoning models handle efficiently with higher accuracy. Our findings underscore the need for more effective training and inference strategies in RLMs.
Ensuring the reliability and verifiability of large language model (LLM)-enabled systems remains a significant challenge in software engineering. We propose a probabilistic framework for systematically analyzing and improving these systems by modeling and refining distributions over clusters of semantically equivalent outputs. This framework facilitates the evaluation and iterative improvement of Transference Models--key software components that utilize LLMs to transform inputs into outputs for downstream tasks. To illustrate its utility, we apply the framework to the autoformalization problem, where natural language documentation is transformed into formal program specifications. Our case illustrates how distribution-aware analysis enables the identification of weaknesses and guides focused alignment improvements, resulting in more reliable and interpretable outputs. This principled approach offers a foundation for addressing critical challenges in the development of robust LLM-enabled systems.
A/B testing experiment is a widely adopted method for evaluating UI/UX design decisions in modern web applications. Yet, traditional A/B testing remains constrained by its dependence on the large-scale and live traffic of human participants, and the long time of waiting for the testing result. Through formative interviews with six experienced industry practitioners, we identified critical bottlenecks in current A/B testing workflows. In response, we present AgentA/B, a novel system that leverages Large Language Model-based autonomous agents (LLM Agents) to automatically simulate user interaction behaviors with real webpages. AgentA/B enables scalable deployment of LLM agents with diverse personas, each capable of navigating the dynamic webpage and interactively executing multi-step interactions like search, clicking, filtering, and purchasing. In a demonstrative controlled experiment, we employ AgentA/B to simulate a between-subject A/B testing with 1,000 LLM agents Amazon.com, and compare agent behaviors with real human shopping behaviors at a scale. Our findings suggest AgentA/B can emulate human-like behavior patterns.
Recent advances in reinforcement learning (RL)-based post-training have led to notable improvements in large language models (LLMs), particularly in enhancing their reasoning capabilities to handle complex tasks. However, most existing methods treat the training data as a unified whole, overlooking the fact that modern LLM training often involves a mixture of data from diverse distributions-varying in both source and difficulty. This heterogeneity introduces a key challenge: how to adaptively schedule training across distributions to optimize learning efficiency. In this paper, we present a principled curriculum learning framework grounded in the notion of distribution-level learnability. Our core insight is that the magnitude of policy advantages reflects how much a model can still benefit from further training on a given distribution. Based on this, we propose a distribution-level curriculum learning framework for RL-based LLM post-training, which leverages the Upper Confidence Bound (UCB) principle to dynamically adjust sampling probabilities for different distrubutions. This approach prioritizes distributions with either high average advantage (exploitation) or low sample count (exploration), yielding an adaptive and theoretically grounded training schedule. We instantiate our curriculum learning framework with GRPO as the underlying RL algorithm and demonstrate its effectiveness on logic reasoning datasets with multiple difficulties and sources. Our experiments show that our framework significantly improves convergence speed and final performance, highlighting the value of distribution-aware curriculum strategies in LLM post-training. Code: https://github.com/ZhentingWang/DUMP.
Developers often evolve an existing software system by making internal changes, called migration. Moving to a new framework, changing implementation to improve efficiency, and upgrading a dependency to its latest version are examples of migrations. Migration is a common and typically continuous maintenance task undertaken either manually or through tooling. Certain migrations are labor intensive and costly, developers do not find the required work rewarding, and they may take years to complete. Hence, automation is preferred for such migrations. In this paper, we discuss a large-scale, costly and traditionally manual migration project at Google, propose a novel automated algorithm that uses change location discovery and a Large Language Model (LLM) to aid developers conduct the migration, report the results of a large case study, and discuss lessons learned. Our case study on 39 distinct migrations undertaken by three developers over twelve months shows that a total of 595 code changes with 93,574 edits have been submitted, where 74.45% of the code changes and 69.46% of the edits were generated by the LLM. The developers reported high satisfaction with the automated tooling, and estimated a 50% reduction on the total time spent on the migration compared to earlier manual migrations. Our results suggest that our automated, LLM-assisted workflow can serve as a model for similar initiatives.
Medication errors significantly threaten patient safety, leading to adverse drug events and substantial economic burdens on healthcare systems. Clinical Decision Support Systems (CDSSs) aimed at mitigating these errors often face limitations when processing unstructured clinical data, including reliance on static databases and rule-based algorithms, frequently generating excessive alerts that lead to alert fatigue among healthcare providers. This paper introduces HELIOT, an innovative CDSS for adverse drug reaction management that processes free-text clinical information using Large Language Models (LLMs) integrated with a comprehensive pharmaceutical data repository. HELIOT leverages advanced natural language processing capabilities to interpret medical narratives, extract relevant drug reaction information from unstructured clinical notes, and learn from past patient-specific medication tolerances to reduce false alerts, enabling more nuanced and contextual adverse drug event warnings across primary care, specialist consultations, and hospital settings. An initial evaluation using a synthetic dataset of clinical narratives and expert-verified ground truth shows promising results. HELIOT achieves high accuracy in a controlled setting. In addition, by intelligently analyzing previous medication tolerance documented in clinical notes and distinguishing between cases requiring different alert types, HELIOT can potentially reduce interruptive alerts by over 50% compared to traditional CDSSs. While these preliminary findings are encouraging, real-world validation will be essential to confirm these benefits in clinical practice.
This paper introduces a novel framework for designing efficient neural network architectures specifically tailored to tiny machine learning (TinyML) platforms. By leveraging large language models (LLMs) for neural architecture search (NAS), a vision transformer (ViT)-based knowledge distillation (KD) strategy, and an explainability module, the approach strikes an optimal balance between accuracy, computational efficiency, and memory usage. The LLM-guided search explores a hierarchical search space, refining candidate architectures through Pareto optimization based on accuracy, multiply-accumulate operations (MACs), and memory metrics. The best-performing architectures are further fine-tuned using logits-based KD with a pre-trained ViT-B/16 model, which enhances generalization without increasing model size. Evaluated on the CIFAR-100 dataset and deployed on an STM32H7 microcontroller (MCU), the three proposed models, LMaNet-Elite, LMaNet-Core, and QwNet-Core, achieve accuracy scores of 74.50%, 74.20% and 73.00%, respectively. All three models surpass current state-of-the-art (SOTA) models, such as MCUNet-in3/in4 (69.62% / 72.86%) and XiNet (72.27%), while maintaining a low computational cost of less than 100 million MACs and adhering to the stringent 320 KB static random-access memory (SRAM) constraint. These results demonstrate the efficiency and performance of the proposed framework for TinyML platforms, underscoring the potential of combining LLM-driven search, Pareto optimization, KD, and explainability to develop accurate, efficient, and interpretable models. This approach opens new possibilities in NAS, enabling the design of efficient architectures specifically suited for TinyML.
Understanding LLM Behaviors via Compression: Data Generation, Knowledge Acquisition and Scaling Laws
Large Language Models (LLMs) have demonstrated remarkable capabilities across numerous tasks, yet principled explanations for their underlying mechanisms and several phenomena, such as scaling laws, hallucinations, and related behaviors, remain elusive. In this work, we revisit the classical relationship between compression and prediction, grounded in Kolmogorov complexity and Shannon information theory, to provide deeper insights into LLM behaviors. By leveraging the Kolmogorov Structure Function and interpreting LLM compression as a two-part coding process, we offer a detailed view of how LLMs acquire and store information across increasing model and data scales -- from pervasive syntactic patterns to progressively rarer knowledge elements. Motivated by this theoretical perspective and natural assumptions inspired by Heap's and Zipf's laws, we introduce a simplified yet representative hierarchical data-generation framework called the Syntax-Knowledge model. Under the Bayesian setting, we show that prediction and compression within this model naturally lead to diverse learning and scaling behaviors of LLMs. In particular, our theoretical analysis offers intuitive and principled explanations for both data and model scaling laws, the dynamics of knowledge acquisition during training and fine-tuning, factual knowledge hallucinations in LLMs. The experimental results validate our theoretical predictions.
Retrieval-Augmented Generation (RAG) has significantly enhanced the factual accuracy and domain adaptability of Large Language Models (LLMs). This advancement has enabled their widespread deployment across sensitive domains such as healthcare, finance, and enterprise applications. RAG mitigates hallucinations by integrating external knowledge, yet introduces privacy risk and security risk, notably data breaching risk and data poisoning risk. While recent studies have explored prompt injection and poisoning attacks, there remains a significant gap in comprehensive research on controlling inbound and outbound query flows to mitigate these threats. In this paper, we propose an AI firewall, ControlNET, designed to safeguard RAG-based LLM systems from these vulnerabilities. ControlNET controls query flows by leveraging activation shift phenomena to detect adversarial queries and mitigate their impact through semantic divergence. We conduct comprehensive experiments on four different benchmark datasets including Msmarco, HotpotQA, FinQA, and MedicalSys using state-of-the-art open source LLMs (Llama3, Vicuna, and Mistral). Our results demonstrate that ControlNET achieves over 0.909 AUROC in detecting and mitigating security threats while preserving system harmlessness. Overall, ControlNET offers an effective, robust, harmless defense mechanism, marking a significant advancement toward the secure deployment of RAG-based LLM systems.
Recent breakthroughs in large Language Models (LLMs) have enabled various generative tasks on a single model. Real-world services (e.g., OpenAI's ChatGPT [27]) powered by an LLM often concurrently support latency-critical requests for interactive applications (e.g., question-answering systems, referred to as real-time or RT requests) and throughput-oriented requests for back-of-house processing (e.g., documents batch processing [28], referred to best-effort or BE requests), with complex hybrid inference workloads to the underlying model. State-of-the-art (SOTA) LLM serving systems dedicate machines to each type of request, towards either low inference latency or high serving throughput, respectively. This practice simplifies request scheduling and management but suffers from poor resource utilization. We propose BROS, a hybrid LLM serving system that aims to collocate RT/BE requests, meeting RT requests' latency requirements while maintaining BE requests' throughput. BROS formulates the problem of hybrid RT/BE request scheduling and solves it with a dynamic priority-based algorithm. BROS designs a bidirectional KV cache management mechanism, allowing RT requests to share KV memory with BE requests to remove the scheduling restrictions caused by insufficient KV memory and improve utilization. Extensive experiments validate that BROS achieves a good trade-off when serving hybrid RT and BE requests. It significantly reduces the latency of RT requests (up to 74.20%), improving their fine-grained service level objectives (SLOs) attainments (up to 36.38x), with negligible throughput reduction for BE requests, showing significant advantages over SOTA systems like vLLM and TGI.
Recent years have witnessed significant progress in large language models' (LLMs) reasoning, which is largely due to the chain-of-thought (CoT) approaches, allowing models to generate intermediate reasoning steps before reaching the final answer. Building on these advances, state-of-the-art LLMs are instruction-tuned to provide long and detailed CoT pathways when responding to reasoning-related questions. However, human beings are naturally cognitive misers and will prompt language models to give rather short responses, thus raising a significant conflict with CoT reasoning. In this paper, we delve into how LLMs' reasoning performance changes when users provide short-path prompts. The results and analysis reveal that language models can reason effectively and robustly without explicit CoT prompts, while under short-path prompting, LLMs' reasoning ability drops significantly and becomes unstable, even on grade-school problems. To address this issue, we propose two approaches: an instruction-guided approach and a fine-tuning approach, both designed to effectively manage the conflict. Experimental results show that both methods achieve high accuracy, providing insights into the trade-off between instruction adherence and reasoning accuracy in current models.
When the complete source sentence is provided, Large Language Models (LLMs) perform excellently in offline machine translation even with a simple prompt "Translate the following sentence from [src lang] into [tgt lang]:". However, in many real scenarios, the source tokens arrive in a streaming manner and simultaneous machine translation (SiMT) is required, then the efficiency and performance of decoder-only LLMs are significantly limited by their auto-regressive nature. To enable LLMs to achieve high-quality SiMT as efficiently as offline translation, we propose a novel paradigm that includes constructing supervised fine-tuning (SFT) data for SiMT, along with new training and inference strategies. To replicate the token input/output stream in SiMT, the source and target tokens are rearranged into an interleaved sequence, separated by special tokens according to varying latency requirements. This enables powerful LLMs to learn read and write operations adaptively, based on varying latency prompts, while still maintaining efficient auto-regressive decoding. Experimental results show that, even with limited SFT data, our approach achieves state-of-the-art performance across various SiMT benchmarks, and preserves the original abilities of offline translation. Moreover, our approach generalizes well to document-level SiMT setting without requiring specific fine-tuning, even beyond the offline translation model.
Chain-of-Thought (CoT) prompting enhances the reasoning of large language models (LLMs) by decomposing problems into sequential steps, mimicking human logic and reducing errors. However, complex tasks with vast solution spaces and vague constraints often exceed the capacity of a single reasoning chain. Inspired by Minimal Free Resolution (MFR) in commutative algebra and algebraic geometry, we propose Syzygy of Thoughts (SoT)-a novel framework that extends CoT by introducing auxiliary, interrelated reasoning paths. SoT captures deeper logical dependencies, enabling more robust and structured problem-solving. MFR decomposes a module into a sequence of free modules with minimal rank, providing a structured analytical approach to complex systems. This method introduces the concepts of "Module", "Betti numbers","Freeness", "Mapping", "Exactness" and "Minimality", enabling the systematic decomposition of the original complex problem into logically complete minimal subproblems while preserving key problem features and reducing reasoning length. We tested SoT across diverse datasets (e.g., GSM8K, MATH) and models (e.g., GPT-4o-mini, Qwen2.5), achieving inference accuracy that matches or surpasses mainstream CoTs standards. Additionally, by aligning the sampling process with algebraic constraints, our approach enhances the scalability of inference time in LLMs, ensuring both transparent reasoning and high performance. Our code will be publicly available at https://github.com/dlMARiA/Syzygy-of-thoughts.
In this paper, we propose LoopLynx, a scalable dataflow architecture for efficient LLM inference that optimizes FPGA usage through a hybrid spatial-temporal design. The design of LoopLynx incorporates a hybrid temporal-spatial architecture, where computationally intensive operators are implemented as large dataflow kernels. This achieves high throughput similar to spatial architecture, and organizing and reusing these kernels in a temporal way together enhances FPGA peak performance. Furthermore, to overcome the resource limitations of a single device, we provide a multi-FPGA distributed architecture that overlaps and hides all data transfers so that the distributed accelerators are fully utilized. By doing so, LoopLynx can be effectively scaled to multiple devices to further explore model parallelism for large-scale LLM inference. Evaluation of GPT-2 model demonstrates that LoopLynx can achieve comparable performance to state-of-the-art single FPGA-based accelerations. In addition, compared to Nvidia A100, our accelerator with a dual-FPGA configuration delivers a 2.52x speed-up in inference latency while consuming only 48.1% of the energy.
This research aims to comprehensively explore building a multimodal foundation model for egocentric video understanding. To achieve this goal, we work on three fronts. First, as there is a lack of QA data for egocentric video understanding, we automatically generate 7M high-quality QA samples for egocentric videos ranging from 30 seconds to one hour long in Ego4D based on human-annotated data. This is one of the largest egocentric QA datasets. Second, we contribute a challenging egocentric QA benchmark with 629 videos and 7,026 questions to evaluate the models' ability in recognizing and memorizing visual details across videos of varying lengths. We introduce a new de-biasing evaluation method to help mitigate the unavoidable language bias present in the models being evaluated. Third, we propose a specialized multimodal architecture featuring a novel "Memory Pointer Prompting" mechanism. This design includes a \textit{global glimpse} step to gain an overarching understanding of the entire video and identify key visual information, followed by a fallback step that utilizes the key visual information to generate responses. This enables the model to more effectively comprehend extended video content. With the data, benchmark, and model, we build MM-Ego, an egocentric multimodal LLM that shows powerful performance on egocentric video understanding.
Large language models learn and continually learn through the accumulation of gradient-based updates, but how individual pieces of new information affect existing knowledge, leading to both beneficial generalization and problematic hallucination, remains poorly understood. We demonstrate that when learning new information, LLMs exhibit a "priming" effect: learning a new fact can cause the model to inappropriately apply that knowledge in unrelated contexts. To systematically study this phenomenon, we introduce "Outlandish," a carefully curated dataset of 1320 diverse text samples designed to probe how new knowledge permeates through an LLM's existing knowledge base. Using this dataset, we show that the degree of priming after learning new information can be predicted by measuring the token probability of key words before learning. This relationship holds robustly across different model architectures (PALM-2, Gemma, Llama), sizes, and training stages. Finally, we develop two novel techniques to modulate how new knowledge affects existing model behavior: (1) a ``stepping-stone'' text augmentation strategy and (2) an ``ignore-k'' update pruning method. These approaches reduce undesirable priming effects by 50-95\% while preserving the model's ability to learn new information. Our findings provide both empirical insights into how LLMs learn and practical tools for improving the specificity of knowledge insertion in language models. Further materials: https://sunchipsster1.github.io/projects/outlandish/
When applying pre-trained large language models (LLMs) to address anomaly detection tasks, the multivariate time series (MTS) modality of anomaly detection does not align with the text modality of LLMs. Existing methods simply transform the MTS data into multiple univariate time series sequences, which can cause many problems. This paper introduces MADLLM, a novel multivariate anomaly detection method via pre-trained LLMs. We design a new triple encoding technique to align the MTS modality with the text modality of LLMs. Specifically, this technique integrates the traditional patch embedding method with two novel embedding approaches: Skip Embedding, which alters the order of patch processing in traditional methods to help LLMs retain knowledge of previous features, and Feature Embedding, which leverages contrastive learning to allow the model to better understand the correlations between different features. Experimental results demonstrate that our method outperforms state-of-the-art methods in various public anomaly detection datasets.
With the rise of large language models (LLMs), researchers are increasingly exploring their applications in var ious vertical domains, such as software engineering. LLMs have achieved remarkable success in areas including code generation and vulnerability detection. However, they also exhibit numerous limitations and shortcomings. LLM-based agents, a novel tech nology with the potential for Artificial General Intelligence (AGI), combine LLMs as the core for decision-making and action-taking, addressing some of the inherent limitations of LLMs such as lack of autonomy and self-improvement. Despite numerous studies and surveys exploring the possibility of using LLMs in software engineering, it lacks a clear distinction between LLMs and LLM based agents. It is still in its early stage for a unified standard and benchmarking to qualify an LLM solution as an LLM-based agent in its domain. In this survey, we broadly investigate the current practice and solutions for LLMs and LLM-based agents for software engineering. In particular we summarise six key topics: requirement engineering, code generation, autonomous decision-making, software design, test generation, and software maintenance. We review and differentiate the work of LLMs and LLM-based agents from these six topics, examining their differences and similarities in tasks, benchmarks, and evaluation metrics. Finally, we discuss the models and benchmarks used, providing a comprehensive analysis of their applications and effectiveness in software engineering. We anticipate this work will shed some lights on pushing the boundaries of LLM-based agents in software engineering for future research.
Large Language Models (LLMs) have recently garnered widespread attention due to their adeptness at generating innovative responses to the given prompts across a multitude of domains. However, LLMs often suffer from the inherent limitation of hallucinations and generate incorrect information while maintaining well-structured and coherent responses. In this work, we hypothesize that hallucinations stem from the internal dynamics of LLMs. Our observations indicate that, during passage generation, LLMs tend to deviate from factual accuracy in subtle parts of responses, eventually shifting toward misinformation. This phenomenon bears a resemblance to human cognition, where individuals may hallucinate while maintaining logical coherence, embedding uncertainty within minor segments of their speech. To investigate this further, we introduce an innovative approach, HalluShift, designed to analyze the distribution shifts in the internal state space and token probabilities of the LLM-generated responses. Our method attains superior performance compared to existing baselines across various benchmark datasets. Our codebase is available at https://github.com/sharanya-dasgupta001/hallushift.
Despite extensive efforts in safety alignment, large language models (LLMs) remain vulnerable to jailbreak attacks. Activation steering offers a training-free defense method but relies on fixed steering coefficients, resulting in suboptimal protection and increased false rejections of benign inputs. To address this, we propose AdaSteer, an adaptive activation steering method that dynamically adjusts model behavior based on input characteristics. We identify two key properties: Rejection Law (R-Law), which shows that stronger steering is needed for jailbreak inputs opposing the rejection direction, and Harmfulness Law (H-Law), which differentiates adversarial and benign inputs. AdaSteer steers input representations along both the Rejection Direction (RD) and Harmfulness Direction (HD), with adaptive coefficients learned via logistic regression, ensuring robust jailbreak defense while preserving benign input handling. Experiments on LLaMA-3.1, Gemma-2, and Qwen2.5 show that AdaSteer outperforms baseline methods across multiple jailbreak attacks with minimal impact on utility. Our results highlight the potential of interpretable model internals for real-time, flexible safety enforcement in LLMs.
As large language models become increasingly capable at various writing tasks, their weakness at generating unique and creative content becomes a major liability. Although LLMs have the ability to generate text covering diverse topics, there is an overall sense of repetitiveness across texts that we aim to formalize and quantify via a similarity metric. The familiarity between documents arises from the persistence of underlying discourse structures. However, existing similarity metrics dependent on lexical overlap and syntactic patterns largely capture $\textit{content}$ overlap, thus making them unsuitable for detecting $\textit{structural}$ similarities. We introduce an abstraction based on linguistic theories in Questions Under Discussion (QUD) and question semantics to help quantify differences in discourse progression. We then use this framework to build $\textbf{QUDsim}$, a similarity metric that can detect discursive parallels between documents. Using QUDsim, we find that LLMs often reuse discourse structures (more so than humans) across samples, even when content differs. Furthermore, LLMs are not only repetitive and structurally uniform, but are also divergent from human authors in the types of structures they use.
Evaluations of large language model (LLM) risks and capabilities are increasingly being incorporated into AI risk management and governance frameworks. Currently, most risk evaluations are conducted by designing inputs that elicit harmful behaviors from the system. However, this approach suffers from two limitations. First, input-output evaluations cannot evaluate realistic risks from open-weight models. Second, the behaviors identified during any particular input-output evaluation can only lower-bound the model's worst-possible-case input-output behavior. As a complementary method for eliciting harmful behaviors, we propose evaluating LLMs with model tampering attacks which allow for modifications to latent activations or weights. We pit state-of-the-art techniques for removing harmful LLM capabilities against a suite of 5 input-space and 6 model tampering attacks. In addition to benchmarking these methods against each other, we show that (1) model resilience to capability elicitation attacks lies on a low-dimensional robustness subspace; (2) the attack success rate of model tampering attacks can empirically predict and offer conservative estimates for the success of held-out input-space attacks; and (3) state-of-the-art unlearning methods can easily be undone within 16 steps of fine-tuning. Together these results highlight the difficulty of suppressing harmful LLM capabilities and show that model tampering attacks enable substantially more rigorous evaluations than input-space attacks alone.
Large Language Models (LLMs) have shown remarkable capabilities across tasks, yet they often require additional prompting techniques when facing complex problems. While approaches like self-correction and response selection have emerged as popular solutions, recent studies have shown these methods perform poorly when relying on the LLM itself to provide feedback or selection criteria. We argue this limitation stems from the fact that common LLM post-training procedures lack explicit supervision for discriminative judgment tasks. In this paper, we propose Generative Self-Aggregation (GSA), a novel prompting method that improves answer quality without requiring the model's discriminative capabilities. GSA first samples multiple diverse responses from the LLM, then aggregates them to obtain an improved solution. Unlike previous approaches, our method does not require the LLM to correct errors or compare response quality; instead, it leverages the model's generative abilities to synthesize a new response based on the context of multiple samples. While GSA shares similarities with the self-consistency (SC) approach for response aggregation, SC requires specific verifiable tokens to enable majority voting. In contrast, our approach is more general and can be applied to open-ended tasks. Empirical evaluation demonstrates that GSA effectively improves response quality across various tasks, including mathematical reasoning, knowledge-based problems, and open-ended generation tasks such as code synthesis and conversational responses.
Mixture of Experts (MoE) LLMs, characterized by their sparse activation patterns, offer a promising approach to scaling language models while avoiding proportionally increasing the inference cost. However, their large parameter sizes present deployment challenges in resource-constrained environments with limited GPU memory capacity, as GPU memory is often insufficient to accommodate the full set of model weights. Consequently, typical deployments rely on CPU-GPU hybrid execution: the GPU handles compute-intensive GEMM operations, while the CPU processes the relatively lightweight attention mechanism. This setup introduces a key challenge: how to effectively optimize resource utilization across CPU and GPU? Prior work has designed system optimizations based on performance models with limited scope. Specifically, such models do not capture the complex interactions between hardware properties and system execution mechanisms. Therefore, previous approaches neither identify nor achieve the hardware limit. This paper presents MoE-Lens, a high-throughput MoE LLM inference system designed through holistic performance modeling for resource-constrained environments. Our performance model thoroughly analyzes various fundamental system components, including CPU memory capacity, GPU compute power, and workload characteristics, to understand the theoretical performance upper bound of MoE inference. Furthermore, it captures the system execution mechanisms to identify the key hardware bottlenecks and accurately predict the achievable throughput. Informed by our performance model, MoE-Lens introduces an inference system approaching hardware limits. Evaluated on diverse MoE models and datasets, MoE-Lens outperforms the state-of-the-art solution by 4.6x on average (up to 25.5x), with our theoretical model predicting performance with an average 94% accuracy.
Training LLMs in distributed environments presents significant challenges due to the complexity of model execution, deployment systems, and the vast space of configurable strategies. Although various optimization techniques exist, achieving high efficiency in practice remains difficult. Accurate performance models that effectively characterize and predict a model's behavior are essential for guiding optimization efforts and system-level studies. We propose Lumos, a trace-driven performance modeling and estimation toolkit for large-scale LLM training, designed to accurately capture and predict the execution behaviors of modern LLMs. We evaluate Lumos on a production ML cluster with up to 512 NVIDIA H100 GPUs using various GPT-3 variants, demonstrating that it can replay execution time with an average error of just 3.3%, along with other runtime details, across different models and configurations. Additionally, we validate its ability to estimate performance for new setups from existing traces, facilitating efficient exploration of model and deployment configurations.
Verifying the provenance of content is crucial to the function of many organizations, e.g., educational institutions, social media platforms, firms, etc. This problem is becoming increasingly challenging as text generated by Large Language Models (LLMs) becomes almost indistinguishable from human-generated content. In addition, many institutions utilize in-house LLMs and want to ensure that external, non-sanctioned LLMs do not produce content within the institution. We answer the following question: Given a piece of text, can we identify whether it was produced by LLM $A$ or $B$ (where $B$ can be a human)? We model LLM-generated text as a sequential stochastic process with complete dependence on history and design zero-shot statistical tests to distinguish between (i) the text generated by two different sets of LLMs $A$ (in-house) and $B$ (non-sanctioned) and also (ii) LLM-generated and human-generated texts. We prove that our tests' type I and type II errors decrease exponentially as text length increases. For designing our tests for a given string, we demonstrate that if the string is generated by the evaluator model $A$, the log-perplexity of the string under $A$ converges to the average entropy of the string under $A$, except with an exponentially small probability in the string length. We also show that if $B$ generates the text, except with an exponentially small probability in string length, the log-perplexity of the string under $A$ converges to the average cross-entropy of $B$ and $A$. For our experiments: First, we present experiments using open-source LLMs to support our theoretical results, and then we provide experiments in a black-box setting with adversarial attacks. Practically, our work enables guaranteed finding of the origin of harmful or false LLM-generated text, which can be useful for combating misinformation and compliance with emerging AI regulations.
Discovering efficient algorithms for solving complex problems has been an outstanding challenge in mathematics and computer science, requiring substantial human expertise over the years. Recent advancements in evolutionary search with large language models (LLMs) have shown promise in accelerating the discovery of algorithms across various domains, particularly in mathematics and optimization. However, existing approaches treat the LLM as a static generator, missing the opportunity to update the model with the signal obtained from evolutionary exploration. In this work, we propose to augment LLM-based evolutionary search by continuously refining the search operator - the LLM - through reinforcement learning (RL) fine-tuning. Our method leverages evolutionary search as an exploration strategy to discover improved algorithms, while RL optimizes the LLM policy based on these discoveries. Our experiments on three combinatorial optimization tasks - bin packing, traveling salesman, and the flatpack problem - show that combining RL and evolutionary search improves discovery efficiency of improved algorithms, showcasing the potential of RL-enhanced evolutionary strategies to assist computer scientists and mathematicians for more efficient algorithm design.
Modern large language model (LLM) serving must efficiently handle highly dynamic workloads, where prompt and response lengths vary significantly across requests. Existing systems typically adopt either colocated execution, where prefill and decode stages share the same GPU for high throughput, or disaggregated execution, which decouples the two stages and assign their tasks to dedicated GPUs for interference avoidance. However, both paradigms face critical limitations: colocation suffers from resource contention and prolonged tail latency, whereas disaggregation likely leads to resource wasting when prefill or decode GPUs are not fully occupied. To address the above limitations, we introduce DynaServe, a unified LLM serving framework based on the Tandem Serving model. Under this model, DynaServe elastically decomposes each request into two virtual sub-requests that are collaboratively processed by a pair of GPU instances. The Lead GPU handles the initial prompt and early generation, while the Follow GPU completes decoding, enabling dynamic load balancing, fine-grained batching, and coherent execution across distributed resources. By coordinating computation and memory across the cluster, DynaServe adapts to diverse and bursty workloads while maintaining stringent latency service-level objectives (SLOs). Evaluations on real-world traces show that DynaServe improves end-to-end Serving Capacity by up to 1.23 $\times$, increases the overall goodput from 1.15 $\times$ to 4.34 $\times$, and improve the memory utilization by up to 49% compared to state-of-the-art colocated and disaggregated systems.
Finding relevant tables among databases, lakes, and repositories is the first step in extracting value from data. Such a task remains difficult because assessing whether a table is relevant to a problem does not always depend only on its content but also on the context, which is usually tribal knowledge known to the individual or team. While tools like data catalogs and academic data discovery systems target this problem, they rely on keyword search or more complex interfaces, limiting non-technical users' ability to find relevant data. The advent of large language models (LLMs) offers a unique opportunity for users to ask questions directly in natural language, making dataset discovery more intuitive, accessible, and efficient. In this paper, we introduce Pneuma, a retrieval-augmented generation (RAG) system designed to efficiently and effectively discover tabular data. Pneuma leverages large language models (LLMs) for both table representation and table retrieval. For table representation, Pneuma preserves schema and row-level information to ensure comprehensive data understanding. For table retrieval, Pneuma augments LLMs with traditional information retrieval techniques, such as full-text and vector search, harnessing the strengths of both to improve retrieval performance. To evaluate Pneuma, we generate comprehensive benchmarks that simulate table discovery workload on six real-world datasets including enterprise data, scientific databases, warehousing data, and open data. Our results demonstrate that Pneuma outperforms widely used table search systems (such as full-text search and state-of-the-art RAG systems) in accuracy and resource efficiency.
Evaluating Personalized Tool-Augmented LLMs from the Perspectives of Personalization and Proactivity
Personalized tool utilization is essential for aligning large language models (LLMs) with user preference in interaction scenarios with various tools. However, most of the current benchmarks primarily focus on either personalization of text generation or direct tool-utilizing, without considering both. In this work, we introduce a novel benchmark ETAPP for evaluating personalized tool invocation, establishing a sandbox environment, and a comprehensive dataset of 800 testing cases covering diverse user profiles. To improve the accuracy of our evaluation, we propose a key-point-based LLM evaluation method, mitigating biases in the LLM-as-a-judge system by manually annotating key points for each test case and providing them to LLM as the reference. Additionally, we evaluate the excellent LLMs and provide an in-depth analysis. Furthermore, we investigate the impact of different tool-invoking strategies on LLMs' personalization performance and the effects of fine-tuning in our task. The effectiveness of our preference-setting and key-point-based evaluation method is also validated. Our findings offer insights into improving personalized LLM agents. Our Code is available at https://github.com/hypasd-art/ETAPP.
Mitigating hallucination issues is a key challenge that must be overcome to reliably deploy large language models (LLMs) in real-world scenarios. Recently, various methods have been proposed to detect and revise factual errors in LLM-generated texts, in order to reduce hallucination. In this paper, we propose Re-Ex, a method for post-editing LLM-generated responses. Re-Ex introduces a novel reasoning step dubbed as the factual error explanation step. Re-Ex revises the initial response of LLMs using 3-steps : first, external tools are used to retrieve the evidences of the factual errors in the initial LLM response; next, LLM is instructed to explain the problematic parts of the response based on the gathered evidence; finally, LLM revises the initial response using the explanations provided in the previous step. In addition to the explanation step, Re-Ex also incorporates new prompting techniques to reduce the token count and inference time required for the response revision process. Compared with existing methods including FacTool, CoVE, and RARR, Re-Ex provides better detection and revision performance with less inference time and fewer tokens in multiple benchmarks.
Emerging services such as augmented reality (AR) and virtual reality (VR) have increased the volume of data transmitted in wireless communication systems, revealing the limitations of traditional Shannon theory. To address these limitations, semantic communication has been proposed as a solution that prioritizes the meaning of messages over the exact transmission of bits. This paper explores semantic communication for text data transmission in end-to-end (E2E) systems through a novel approach called KG-LLM semantic communication, which integrates knowledge graph (KG) extraction and large language model (LLM) coding. In this method, the transmitter first utilizes a KG to extract key entities and relationships from sentences. The extracted information is then encoded using an LLM to obtain the semantic meaning. On the receiver side, messages are decoded using another LLM, while a bidirectional encoder representations from transformers (i.e., BERT) model further refines the reconstructed sentences for improved semantic similarity. The KG-LLM semantic communication method reduces the transmitted text data volume by 30% through KG-based compression and achieves 84\% semantic similarity between the original and received messages. This demonstrates the KG-LLM methods efficiency and robustness in semantic communication systems, outperforming the deep learning-based semantic communication model (DeepSC), which achieves only 63%.
Enhancing the intelligence of smart systems, such as smart home, and smart vehicle, and smart grids, critically depends on developing sophisticated planning capabilities that can anticipate the next desired function based on historical interactions. While existing methods view user behaviors as sequential data and apply models like RNNs and Transformers to predict future actions, they often fail to incorporate domain knowledge and capture personalized user preferences. In this paper, we propose a novel approach that incorporates LLM-enhanced logs and personalized prompts. Our approach first constructs a graph that captures individual behavior preferences derived from their interaction histories. This graph effectively transforms into a soft continuous prompt that precedes the sequence of user behaviors. Then our approach leverages the vast general knowledge and robust reasoning capabilities of a pretrained LLM to enrich the oversimplified and incomplete log records. By enhancing these logs semantically, our approach better understands the user's actions and intentions, especially for those rare events in the dataset. We evaluate the method across four real-world datasets from both smart vehicle and smart home settings. The findings validate the effectiveness of our LLM-enhanced description and personalized prompt, shedding light on potential ways to advance the intelligence of smart space.
Advances in Large Language Models (LLMs) have significantly improved multi-step reasoning through generating free-text rationales. However, recent studies show that LLMs tend to lose focus over the middle of long contexts. This raises concerns that as reasoning progresses, LLMs may overlook information in earlier steps when decoding subsequent steps, leading to generate unreliable and redundant rationales. To address this, we propose guiding LLMs to generate more accurate and concise step-by-step rationales by (1) proactively referencing information from underutilized prior steps, and (2) minimizing redundant information between new and existing steps. We introduce stepwise informativeness search, an inference-time tree search framework incorporating two selection heuristics: grounding-guided selection which prioritizes steps paying higher attention over underutilized steps; and novelty-guided selection which encourages steps with novel conclusions. During rationale generation, we use a self-grounding strategy that prompts LLMs to explicitly reference relevant prior steps to provide premises before deduction at each step. Experimental results on four reasoning datasets demonstrate that our approach improves reasoning accuracy by generating higher-quality rationales with reduced errors and redundancy.
Comedy serves as a profound reflection of the times we live in and is a staple element of human interactions. In light of the widespread adoption of Large Language Models (LLMs), the intersection of humor and AI has become no laughing matter. Advancements in the naturalness of human-computer interaction correlates with improvements in AI systems' abilities to understand humor. In this study, we assess the ability of models in accurately identifying humorous quotes from a stand-up comedy transcript. Stand-up comedy's unique comedic narratives make it an ideal dataset to improve the overall naturalness of comedic understanding. We propose a novel humor detection metric designed to evaluate LLMs amongst various prompts on their capability to extract humorous punchlines. The metric has a modular structure that offers three different scoring methods - fuzzy string matching, sentence embedding, and subspace similarity - to provide an overarching assessment of a model's performance. The model's results are compared against those of human evaluators on the same task. Our metric reveals that regardless of prompt engineering, leading models, ChatGPT, Claude, and DeepSeek, achieve scores of at most 51% in humor detection. Notably, this performance surpasses that of humans who achieve a score of 41%. The analysis of human evaluators and LLMs reveals variability in agreement, highlighting the subjectivity inherent in humor and the complexities involved in extracting humorous quotes from live performance transcripts. Code available at https://github.com/swaggirl9000/humor.
Reasoning is a fundamental cognitive process that enables logical inference, problem-solving, and decision-making. With the rapid advancement of large language models (LLMs), reasoning has emerged as a key capability that distinguishes advanced AI systems from conventional models that empower chatbots. In this survey, we categorize existing methods along two orthogonal dimensions: (1) Regimes, which define the stage at which reasoning is achieved (either at inference time or through dedicated training); and (2) Architectures, which determine the components involved in the reasoning process, distinguishing between standalone LLMs and agentic compound systems that incorporate external tools, and multi-agent collaborations. Within each dimension, we analyze two key perspectives: (1) Input level, which focuses on techniques that construct high-quality prompts that the LLM condition on; and (2) Output level, which methods that refine multiple sampled candidates to enhance reasoning quality. This categorization provides a systematic understanding of the evolving landscape of LLM reasoning, highlighting emerging trends such as the shift from inference-scaling to learning-to-reason (e.g., DeepSeek-R1), and the transition to agentic workflows (e.g., OpenAI Deep Research, Manus Agent). Additionally, we cover a broad spectrum of learning algorithms, from supervised fine-tuning to reinforcement learning such as PPO and GRPO, and the training of reasoners and verifiers. We also examine key designs of agentic workflows, from established patterns like generator-evaluator and LLM debate to recent innovations. ...
We present PLAYER*, a novel framework for Large Language Model (LLM)-based agents in Murder Mystery Games (MMGs). MMGs pose unique challenges, including undefined state spaces, absent intermediate rewards, and the need for strategic interaction in a continuous language domain. PLAYER* addresses these complexities through a sensor-based representation of agent states, a question-targeting mechanism guided by information gain, and a pruning strategy to refine suspect lists and enhance decision-making efficiency. To enable systematic evaluation, we propose WellPlay, a dataset comprising 1,482 inferential questions across 12 games, categorized into objectives, reasoning, and relationships. Experiments demonstrate PLAYER*'s capacity to achieve superior performance in reasoning accuracy and efficiency compared to existing approaches, while also significantly improving the quality of agent-human interactions in MMGs. This study advances the development of reasoning agents for complex social and interactive scenarios.
This paper presents AutoEval, a novel benchmark for scaling Large Language Model (LLM) assessment in formal tasks with clear notions of correctness, such as truth maintenance in translation and logical reasoning. AutoEval is the first benchmarking paradigm that offers several key advantages necessary for scaling objective evaluation of LLMs without human labeling: (a) ability to evaluate LLMs of increasing sophistication by auto-generating tasks at different levels of difficulty; (b) auto-generation of ground truth that eliminates dependence on expensive and time-consuming human annotation; (c) the use of automatically generated, randomized datasets that mitigate the ability of successive LLMs to overfit to static datasets used in many contemporary benchmarks. Empirical analysis shows that an LLM's performance on AutoEval is highly indicative of its performance on a diverse array of other benchmarks focusing on translation and reasoning tasks, making it a valuable autonomous evaluation paradigm in settings where hand-curated datasets can be hard to obtain and/or update.
Optimal hyperparameter selection is critical for maximizing neural network performance, especially as models grow in complexity. This work investigates the viability of leveraging large language models (LLMs) for hyperparameter optimization by fine-tuning a parameter-efficient version of Code Llama using LoRA. The adapted LLM is capable of generating accurate and efficient hyperparameter recommendations tailored to diverse neural network architectures. Unlike traditional approaches such as Optuna, which rely on computationally intensive trial-and-error procedures, our method achieves competitive or superior results in terms of Root Mean Square Error (RMSE) while significantly reducing computational overhead. Our findings demonstrate that LLM-based optimization not only matches the performance of state-of-the-art techniques like Tree-structured Parzen Estimators (TPE) but also substantially accelerates the tuning process. This positions LLMs as a promising alternative for rapid experimentation, particularly in resource-constrained environments such as edge devices and mobile platforms, where computational efficiency is essential. In addition to improved efficiency, the method offers time savings and consistent performance across various tasks, highlighting its robustness and generalizability. All generated hyperparameters are included in the LEMUR Neural Network (NN) Dataset, which is publicly available and serves as an open-source benchmark for hyperparameter optimization research.
To complete an open-ended programming exercise, students need to both plan a high-level solution and implement it using the appropriate syntax. However, these problems are often autograded on the correctness of the final submission through test cases, and students cannot get feedback on their planning process. Large language models (LLM) may be able to generate this feedback by detecting the overall code structure even for submissions with syntax errors. To this end, we propose an approach that detects which high-level goals and patterns (i.e. programming plans) exist in a student program with LLMs. We show that both the full GPT-4o model and a small variant (GPT-4o-mini) can detect these plans with remarkable accuracy, outperforming baselines inspired by conventional approaches to code analysis. We further show that the smaller, cost-effective variant (GPT-4o-mini) achieves results on par with state-of-the-art (GPT-4o) after fine-tuning, creating promising implications for smaller models for real-time grading. These smaller models can be incorporated into autograders for open-ended code-writing exercises to provide feedback for students' implicit planning skills, even when their program is syntactically incorrect. Furthermore, LLMs may be useful in providing feedback for problems in other domains where students start with a set of high-level solution steps and iteratively compute the output, such as math and physics problems.
The array of emergent capabilities of large language models (LLMs) has sparked interest in assessing their ability to simulate human opinions in a variety of contexts, potentially serving as surrogates for human subjects in opinion surveys. However, previous evaluations of this capability have depended heavily on costly, domain-specific human survey data, and mixed empirical results about LLM effectiveness create uncertainty for managers about whether investing in this technology is justified in early-stage research. To address these challenges, we introduce a series of quality checks to support early-stage deliberation about the viability of using LLMs for simulating human opinions. These checks emphasize logical constraints, model stability, and alignment with stakeholder expectations of model outputs, thereby reducing dependence on human-generated data in the initial stages of evaluation. We demonstrate the usefulness of the proposed quality control tests in the context of AI-assisted content moderation, an application that both advocates and critics of LLMs' capabilities to simulate human opinion see as a desirable potential use case. None of the tested models passed all quality control checks, revealing several failure modes. We conclude by discussing implications of these failure modes and recommend how organizations can utilize our proposed tests for prompt engineering and in their risk management practices when considering the use of LLMs for opinion simulation. We make our crowdsourced dataset of claims with human and LLM annotations publicly available for future research.
Large Language Models (LLMs) have been shown to demonstrate imbalanced biases against certain groups. However, the study of unprovoked targeted attacks by LLMs towards at-risk populations remains underexplored. Our paper presents three novel contributions: (1) the explicit evaluation of LLM-generated attacks on highly vulnerable mental health groups; (2) a network-based framework to study the propagation of relative biases; and (3) an assessment of the relative degree of stigmatization that emerges from these attacks. Our analysis of a recently released large-scale bias audit dataset reveals that mental health entities occupy central positions within attack narrative networks, as revealed by a significantly higher mean centrality of closeness (p-value = 4.06e-10) and dense clustering (Gini coefficient = 0.7). Drawing from sociological foundations of stigmatization theory, our stigmatization analysis indicates increased labeling components for mental health disorder-related targets relative to initial targets in generation chains. Taken together, these insights shed light on the structural predilections of large language models to heighten harmful discourse and highlight the need for suitable approaches for mitigation.
Code LLMs have shown promising results with converting tasks in natural language to programs that can be executed by service robots. We are interested in finetuning small, specialized LLMs for this purpose, but collecting datasets of task-program pairs specific to each robot is time-consuming and expensive. While approaches such as SELF-INSTRUCT and EVOL-INSTRUCT are capable of generating novel tasks given a few examples, they are unable to provide the corresponding programs that correctly abide by physical-world and robot-constraints using the provided programming interface. Using a simulator is a natural potential solution to checking for such constraints, but building simulation environments that can handle arbitrary tasks and their necessary objects and locations, is challenging. To address these challenges, we introduce ROBO-INSTRUCT, which synthesizes task-specific simulation environments on the fly during program execution, by opportunistically inferring entity properties and enforcing corresponding constraints based on how the entities are used in the task program. Additionally, ROBO-INSTRUCT integrates an LLM-aided post-processing procedure to refine instructions for better alignment with robot programs. We demonstrate the effectiveness of ROBO-INSTRUCT across multiple LLMs, showing that our fine-tuned models outperform all baseline methods and even match or surpass the performance of several larger and proprietary models.
Recent advancements in Large Language Models (LLMs) have transformed communication, yet their role in secure messaging remains underexplored, especially in surveillance-heavy environments. At the same time, many governments all over the world are proposing legislation to detect, backdoor, or even ban encrypted communication. That emphasizes the need for alternative ways to communicate securely and covertly over open channels. We propose a novel cryptographic embedding framework that enables covert Public Key or Symmetric Key encrypted communication over public chat channels with humanlike produced texts. Some unique properties of our framework are: 1. It is LLM agnostic, i.e., it allows participants to use different local LLM models independently; 2. It is pre- or post-quantum agnostic; 3. It ensures indistinguishability from human-like chat-produced texts. Thus, it offers a viable alternative where traditional encryption is detectable and restricted.
LLMs as intelligent agents are being increasingly applied in scenarios where human interactions are involved, leading to a critical concern about whether LLMs are faithful to the variations in culture across regions. Several works have investigated this question in various ways, finding that there are biases present in the cultural representations of LLM outputs. To gain a more comprehensive view, in this work, we conduct the first large-scale evaluation of LLM culture assessing 20 countries' cultures and languages across ten LLMs. With a renowned cultural values questionnaire and by carefully analyzing LLM output with human ground truth scores, we thoroughly study LLMs' cultural alignment across countries and among individual models. Our findings show that the output over all models represents a moderate cultural middle ground. Given the overall skew, we propose an alignment metric, revealing that the United States is the best-aligned country and GLM-4 has the best ability to align to cultural values. Deeper investigation sheds light on the influence of model origin, prompt language, and value dimensions on cultural output. Specifically, models, regardless of where they originate, align better with the US than they do with China. The conclusions provide insight to how LLMs can be better aligned to various cultures as well as provoke further discussion of the potential for LLMs to propagate cultural bias and the need for more culturally adaptable models.
We introduce AI University (AI-U), a flexible framework for AI-driven course content delivery that adapts to instructors' teaching styles. At its core, AI-U fine-tunes a large language model (LLM) with retrieval-augmented generation (RAG) to generate instructor-aligned responses from lecture videos, notes, and textbooks. Using a graduate-level finite-element-method (FEM) course as a case study, we present a scalable pipeline to systematically construct training data, fine-tune an open-source LLM with Low-Rank Adaptation (LoRA), and optimize its responses through RAG-based synthesis. Our evaluation - combining cosine similarity, LLM-based assessment, and expert review - demonstrates strong alignment with course materials. We also have developed a prototype web application, available at https://my-ai-university.com, that enhances traceability by linking AI-generated responses to specific sections of the relevant course material and time-stamped instances of the open-access video lectures. Our expert model is found to have greater cosine similarity with a reference on 86% of test cases. An LLM judge also found our expert model to outperform the base Llama 3.2 model approximately four times out of five. AI-U offers a scalable approach to AI-assisted education, paving the way for broader adoption in higher education. Here, our framework has been presented in the setting of a class on FEM - a subject that is central to training PhD and Master students in engineering science. However, this setting is a particular instance of a broader context: fine-tuning LLMs to research content in science.
We present a system using Multimodal LLMs (MLLMs) to analyze a large database with tens of millions of images captured at different times, with the aim of discovering patterns in temporal changes. Specifically, we aim to capture frequent co-occurring changes ("trends") across a city over a certain period. Unlike previous visual analyses, our analysis answers open-ended queries (e.g., "what are the frequent types of changes in the city?") without any predetermined target subjects or training labels. These properties cast prior learning-based or unsupervised visual analysis tools unsuitable. We identify MLLMs as a novel tool for their open-ended semantic understanding capabilities. Yet, our datasets are four orders of magnitude too large for an MLLM to ingest as context. So we introduce a bottom-up procedure that decomposes the massive visual analysis problem into more tractable sub-problems. We carefully design MLLM-based solutions to each sub-problem. During experiments and ablation studies with our system, we find it significantly outperforms baselines and is able to discover interesting trends from images captured in large cities (e.g., "addition of outdoor dining,", "overpass was painted blue," etc.). See more results and interactive demos at https://boyangdeng.com/visual-chronicles.
To reduce development overhead and enable seamless integration between potential components comprising any given generative AI application, the Model Context Protocol (MCP) (Anthropic, 2024) has recently been released and subsequently widely adopted. The MCP is an open protocol that standardizes API calls to large language models (LLMs), data sources, and agentic tools. By connecting multiple MCP servers, each defined with a set of tools, resources, and prompts, users are able to define automated workflows fully driven by LLMs. However, we show that the current MCP design carries a wide range of security risks for end users. In particular, we demonstrate that industry-leading LLMs may be coerced into using MCP tools to compromise an AI developer's system through various attacks, such as malicious code execution, remote access control, and credential theft. To proactively mitigate these and related attacks, we introduce a safety auditing tool, MCPSafetyScanner, the first agentic tool to assess the security of an arbitrary MCP server. MCPScanner uses several agents to (a) automatically determine adversarial samples given an MCP server's tools and resources; (b) search for related vulnerabilities and remediations based on those samples; and (c) generate a security report detailing all findings. Our work highlights serious security issues with general-purpose agentic workflows while also providing a proactive tool to audit MCP server safety and address detected vulnerabilities before deployment. The described MCP server auditing tool, MCPSafetyScanner, is freely available at: https://github.com/johnhalloran321/mcpSafetyScanner
A widely used Agile practice for requirements is to produce a set of user stories (also called ``agile product backlog''), which roughly includes a list of pairs (role, feature), where the role handles the feature for a certain purpose. In the context of Software Product Lines, the requirements for a family of similar systems is thus a family of user-story sets, one per system, leading to a 3-dimensional dataset composed of sets of triples (system, role, feature). In this paper, we combine Triadic Concept Analysis (TCA) and Large Language Model (LLM) prompting to suggest the user-story set required to develop a new system relying on the variability logic of an existing system family. This process consists in 1) computing 3-dimensional variability expressed as a set of TCA implications, 2) providing the designer with intelligible design options, 3) capturing the designer's selection of options, 4) proposing a first user-story set corresponding to this selection, 5) consolidating its validity according to the implications identified in step 1, while completing it if necessary, and 6) leveraging LLM to have a more comprehensive website. This process is evaluated with a dataset comprising the user-story sets of 67 similar-purpose websites.
There is general agreement that fostering trust and cooperation within the AI development ecosystem is essential to promote the adoption of trustworthy AI systems. By embedding Large Language Model (LLM) agents within an evolutionary game-theoretic framework, this paper investigates the complex interplay between AI developers, regulators and users, modelling their strategic choices under different regulatory scenarios. Evolutionary game theory (EGT) is used to quantitatively model the dilemmas faced by each actor, and LLMs provide additional degrees of complexity and nuances and enable repeated games and incorporation of personality traits. Our research identifies emerging behaviours of strategic AI agents, which tend to adopt more "pessimistic" (not trusting and defective) stances than pure game-theoretic agents. We observe that, in case of full trust by users, incentives are effective to promote effective regulation; however, conditional trust may deteriorate the "social pact". Establishing a virtuous feedback between users' trust and regulators' reputation thus appears to be key to nudge developers towards creating safe AI. However, the level at which this trust emerges may depend on the specific LLM used for testing. Our results thus provide guidance for AI regulation systems, and help predict the outcome of strategic LLM agents, should they be used to aid regulation itself.
In this paper, we examine the use of Conformal Language Modelling (CLM) alongside Answer Set Programming (ASP) to enhance the performance of standard open-weight LLMs on complex multi-step reasoning tasks. Using the StepGame dataset, which requires spatial reasoning, we apply CLM to generate sets of ASP programs from an LLM, providing statistical guarantees on the correctness of the outputs. Experimental results show that CLM significantly outperforms baseline models that use standard sampling methods, achieving substantial accuracy improvements across different levels of reasoning complexity. Additionally, the LLM-as-Judge metric enhances CLM's performance, especially in assessing structurally and logically correct ASP outputs. However, calibrating CLM with diverse calibration sets did not improve generalizability for tasks requiring much longer reasoning steps, indicating limitations in handling more complex tasks.
The Finite Element Method (FEM) is widely used in engineering and scientific computing, but its pre-processing, solver configuration, and post-processing stages are often time-consuming and require specialized knowledge. This paper proposes an automated solution framework, MooseAgent, for the multi-physics simulation framework MOOSE, which combines large-scale pre-trained language models (LLMs) with a multi-agent system. The framework uses LLMs to understand user-described simulation requirements in natural language and employs task decomposition and multi-round iterative verification strategies to automatically generate MOOSE input files. To improve accuracy and reduce model hallucinations, the system builds and utilizes a vector database containing annotated MOOSE input cards and function documentation. We conducted experimental evaluations on several typical cases, including heat transfer, mechanics, phase field, and multi-physics coupling. The results show that MooseAgent can automate the MOOSE simulation process to a certain extent, especially demonstrating a high success rate when dealing with relatively simple single-physics problems. The main contribution of this research is the proposal of a multi-agent automated framework for MOOSE, which validates its potential in simplifying finite element simulation processes and lowering the user barrier, providing new ideas for the development of intelligent finite element simulation software. The code for the MooseAgent framework proposed in this paper has been open-sourced and is available at https://github.com/taozhan18/MooseAgent
Large Language Models (LLMs) are reshaping the landscape of computer science research, driving significant shifts in research priorities across diverse conferences and fields. This study provides a comprehensive analysis of the publication trend of LLM-related papers in 77 top-tier computer science conferences over the past six years (2019-2024). We approach this analysis from four distinct perspectives: (1) We investigate how LLM research is driving topic shifts within major conferences. (2) We adopt a topic modeling approach to identify various areas of LLM-related topic growth and reveal the topics of concern at different conferences. (3) We explore distinct contribution patterns of academic and industrial institutions. (4) We study the influence of national origins on LLM development trajectories. Synthesizing the findings from these diverse analytical angles, we derive ten key insights that illuminate the dynamics and evolution of the LLM research ecosystem.
Large Language Models (LLMs) are increasingly used as autonomous agents for multi-step tasks. However, most existing frameworks fail to maintain a structured understanding of the task state, often relying on linear prompt concatenation or shallow memory buffers. This leads to brittle performance, frequent hallucinations, and poor long-range coherence. In this work, we propose the Task Memory Engine (TME), a lightweight and structured memory module that tracks task execution using a hierarchical Task Memory Tree (TMT). Each node in the tree corresponds to a task step, storing relevant input, output, status, and sub-task relationships. We introduce a prompt synthesis method that dynamically generates LLM prompts based on the active node path, significantly improving execution consistency and contextual grounding. Through case studies and comparative experiments on multi-step agent tasks, we demonstrate that TME leads to better task completion accuracy and more interpretable behavior with minimal implementation overhead. The full implementation of TME is available at https://github.com/biubiutomato/TME-Agent.