llm - 2025_01
Navigation
Papers
We propose affine concept editing (ACE) as an approach for steering language models' behavior by intervening directly in activations. We begin with an affine decomposition of model activation vectors and show that prior methods for steering model behavior correspond to subsets of terms of this decomposition. We then provide a derivation of ACE and use it to control refusal behavior on ten different models, including Llama 3 70B. ACE combines affine subspace projection and activation addition to reliably control the model's refusal responses across prompt types. We evaluate the results using LLM-based scoring on a collection of harmful and harmless prompts. Our experiments demonstrate that ACE consistently achieves more precise control over model behavior than existing methods and generalizes to models where directional ablation via affine subspace projection alone produces incoherent outputs. Code for reproducing our results is available at https://github.com/EleutherAI/steering-llama3 .
Methods to ensure factual accuracy of text generated by large language models (LLM) in clinical medicine are lacking. VeriFact is an artificial intelligence system that combines retrieval-augmented generation and LLM-as-a-Judge to verify whether LLM-generated text is factually supported by a patient's medical history based on their electronic health record (EHR). To evaluate this system, we introduce VeriFact-BHC, a new dataset that decomposes Brief Hospital Course narratives from discharge summaries into a set of simple statements with clinician annotations for whether each statement is supported by the patient's EHR clinical notes. Whereas highest agreement between clinicians was 88.5%, VeriFact achieves up to 92.7% agreement when compared to a denoised and adjudicated average human clinican ground truth, suggesting that VeriFact exceeds the average clinician's ability to fact-check text against a patient's medical record. VeriFact may accelerate the development of LLM-based EHR applications by removing current evaluation bottlenecks.
Mining and conveying actionable insights from complex data is a key challenge of exploratory data analysis (EDA) and storytelling. To address this challenge, we present a design space for actionable EDA and storytelling. Synthesizing theory and expert interviews, we highlight how semantic precision, rhetorical persuasion, and pragmatic relevance underpin effective EDA and storytelling. We also show how this design space subsumes common challenges in actionable EDA and storytelling, such as identifying appropriate analytical strategies and leveraging relevant domain knowledge. Building on the potential of LLMs to generate coherent narratives with commonsense reasoning, we contribute Jupybara, an AI-enabled assistant for actionable EDA and storytelling implemented as a Jupyter Notebook extension. Jupybara employs two strategies -- design-space-aware prompting and multi-agent architectures -- to operationalize our design space. An expert evaluation confirms Jupybara's usability, steerability, explainability, and reparability, as well as the effectiveness of our strategies in operationalizing the design space framework with LLMs.
We present a modular pipeline that automates the generation of stealthy jailbreak prompts derived from high-level content policies, enhancing LLM content moderation. First, we address query inefficiency and jailbreak strength by developing Graph of Attacks with Pruning (GAP), a method that utilizes strategies from prior jailbreaks, resulting in 92% attack success rate on GPT-3.5 using only 54% of the queries of the prior algorithm. Second, we address the cold-start issue by automatically generating seed prompts from the high-level policy using LLMs. Finally, we demonstrate the utility of these generated jailbreak prompts of improving content moderation by fine-tuning PromptGuard, a model trained to detect jailbreaks, increasing its accuracy on the Toxic-Chat dataset from 5.1% to 93.89%.
We present a novel mission-planning strategy for heterogeneous multi-robot teams, taking into account the specific constraints and capabilities of each robot. Our approach employs hierarchical trees to systematically break down complex missions into manageable sub-tasks. We develop specialized APIs and tools, which are utilized by Large Language Models (LLMs) to efficiently construct these hierarchical trees. Once the hierarchical tree is generated, it is further decomposed to create optimized schedules for each robot, ensuring adherence to their individual constraints and capabilities. We demonstrate the effectiveness of our framework through detailed examples covering a wide range of missions, showcasing its flexibility and scalability.
Alignment in large language models (LLMs) is used to enforce guidelines such as safety. Yet, alignment fails in the face of jailbreak attacks that modify inputs to induce unsafe outputs. In this paper, we present and evaluate a method to assess the robustness of LLM alignment. We observe that alignment embeds a safety classifier in the target model that is responsible for deciding between refusal and compliance. We seek to extract an approximation of this classifier, called a surrogate classifier, from the LLM. We develop an algorithm for identifying candidate classifiers from subsets of the LLM model. We evaluate the degree to which the candidate classifiers approximate the model's embedded classifier in benign (F1 score) and adversarial (using surrogates in a white-box attack) settings. Our evaluation shows that the best candidates achieve accurate agreement (an F1 score above 80%) using as little as 20% of the model architecture. Further, we find attacks mounted on the surrogate models can be transferred with high accuracy. For example, a surrogate using only 50% of the Llama 2 model achieved an attack success rate (ASR) of 70%, a substantial improvement over attacking the LLM directly, where we only observed a 22% ASR. These results show that extracting surrogate classifiers is a viable (and highly effective) means for modeling (and therein addressing) the vulnerability of aligned models to jailbreaking attacks.
Large Language Models (LLMs) have become state-of-the-art in Machine Translation (MT), often trained on massive bilingual parallel corpora scraped from the web, that contain low-quality entries and redundant information, leading to significant computational challenges. Various data filtering methods exist to reduce dataset sizes, but their effectiveness largely varies based on specific language pairs and domains. This paper evaluates the impact of commonly used data filtering techniques, such as LASER, MUSE, and LaBSE, on English-Polish translation within the biomedical domain. By filtering the UFAL Medical Corpus, we created varying dataset sizes to fine-tune the mBART50 model, which was then evaluated using the SacreBLEU metric on the Khresmoi dataset, having the quality of translations assessed by bilingual speakers. Our results show that both LASER and MUSE can significantly reduce dataset sizes while maintaining or even enhancing performance. We recommend the use of LASER, as it consistently outperforms the other methods and provides the most fluent and natural-sounding translations.
This research assesses the effectiveness of state-of-the-art large language models (LLMs), including ChatGPT, Llama, Aya, Jais, and ACEGPT, in the task of Arabic automated essay scoring (AES) using the AR-AES dataset. It explores various evaluation methodologies, including zero-shot, few-shot in-context learning, and fine-tuning, and examines the influence of instruction-following capabilities through the inclusion of marking guidelines within the prompts. A mixed-language prompting strategy, integrating English prompts with Arabic content, was implemented to improve model comprehension and performance. Among the models tested, ACEGPT demonstrated the strongest performance across the dataset, achieving a Quadratic Weighted Kappa (QWK) of 0.67, but was outperformed by a smaller BERT-based model with a QWK of 0.88. The study identifies challenges faced by LLMs in processing Arabic, including tokenization complexities and higher computational demands. Performance variation across different courses underscores the need for adaptive models capable of handling diverse assessment formats and highlights the positive impact of effective prompt engineering on improving LLM outputs. To the best of our knowledge, this study is the first to empirically evaluate the performance of multiple generative Large Language Models (LLMs) on Arabic essays using authentic student data.
The rapid development of large language models (LLMs) has necessitated the creation of benchmarks to evaluate their performance. These benchmarks resemble human tests and surveys, as they consist of sets of questions designed to measure emergent properties in the cognitive behavior of these systems. However, unlike the well-defined traits and abilities studied in social sciences, the properties measured by these benchmarks are often vaguer and less rigorously defined. The most prominent benchmarks are often grouped into leaderboards for convenience, aggregating performance metrics and enabling comparisons between models. Unfortunately, these leaderboards typically rely on simplistic aggregation methods, such as taking the average score across benchmarks. In this paper, we demonstrate the advantages of applying contemporary psychometric methodologies - originally developed for human tests and surveys - to improve the ranking of large language models on leaderboards. Using data from the Hugging Face Leaderboard as an example, we compare the results of the conventional naive ranking approach with a psychometrically informed ranking. The findings highlight the benefits of adopting psychometric techniques for more robust and meaningful evaluation of LLM performance.
Large language models (LLMs) have significantly improved their ability to perform tasks in the field of code generation. However, there is still a gap between LLMs being capable coders and being top-tier software engineers. Based on the observation that top-level software engineers often ask clarifying questions to reduce ambiguity in both requirements and coding solutions, we argue that the same should be applied to LLMs for code generation tasks. In this work, we conducted an empirical study on the benchmark and analysis of the communication skills of LLMs for code generation. We define communication skills of LLMs as ``being able to ask clarifying questions when the description of the code generation problem has issues''. We created a new benchmark, HumanEvalComm, by modifying problem descriptions according to three issues: inconsistency, ambiguity, incompleteness. We defined new evaluation metrics such as Communication Rate and Good Question Rate, and then experimented on HumanEvalComm with different Code LLMs, and a new LLM agent approach, Okanagan, to identify and ask questions in ambiguous parts from code and descriptions for further refining the generated code. Finally, we discussed evaluation results by comparing Code LLMs and Okanagan with our findings.
Large Language Models (LLMs) have found widespread applications in various domains, including web applications, where they facilitate human interaction via chatbots with natural language interfaces. Internally, aided by an LLM-integration middleware such as Langchain, user prompts are translated into SQL queries used by the LLM to provide meaningful responses to users. However, unsanitized user prompts can lead to SQL injection attacks, potentially compromising the security of the database. Despite the growing interest in prompt injection vulnerabilities targeting LLMs, the specific risks of generating SQL injection attacks through prompt injections have not been extensively studied. In this paper, we present a comprehensive examination of prompt-to-SQL (P$_2$SQL) injections targeting web applications based on the Langchain framework. Using Langchain as our case study, we characterize P$_2$SQL injections, exploring their variants and impact on application security through multiple concrete examples. Furthermore, we evaluate 7 state-of-the-art LLMs, demonstrating the pervasiveness of P$_2$SQL attacks across language models. Our findings indicate that LLM-integrated applications based on Langchain are highly susceptible to P$_2$SQL injection attacks, warranting the adoption of robust defenses. To counter these attacks, we propose four effective defense techniques that can be integrated as extensions to the Langchain framework. We validate the defenses through an experimental evaluation with a real-world use case application.
Reinforcement learning with human feedback (RLHF) fine-tunes a pretrained large language model (LLM) using user preference data, enabling it to generate content aligned with human preferences. However, due to privacy concerns, users may be reluctant to share sensitive preference data. To address this, we propose utilizing Federated Learning (FL) techniques, allowing large-scale preference collection from diverse real-world users without requiring them to transmit data to a central server. Our federated RLHF methods (i.e., FedBis and FedBiscuit) encode each client's preferences into binary selectors and aggregate them to capture common preferences. In particular, FedBiscuit overcomes key challenges, such as preference heterogeneity and reward hacking, through innovative solutions like grouping clients with similar preferences to reduce heterogeneity and using multiple binary selectors to enhance LLM output quality. To evaluate the performance of the proposed methods, we establish the first federated RLHF benchmark with a heterogeneous human preference dataset. Experimental results show that by integrating the LLM with aggregated client preferences, FedBis and FedBiscuit significantly enhance the professionalism and readability of the generated content.
LLMs have shown preliminary promise in some security tasks and CTF challenges. However, it is unclear whether LLMs are able to realize multistage network attacks, which involve executing a wide variety of actions across multiple hosts such as conducting reconnaissance, exploiting vulnerabilities to gain initial access, leveraging internal hosts to move laterally, and using multiple compromised hosts to exfiltrate data. We evaluate LLMs across 10 multistage networks and find that popular LLMs are unable to realize these attacks. To enable LLMs to realize these attacks, we introduce Incalmo, an LLM-agnostic high-level attack abstraction layer that sits between an LLM and the environment. Rather than LLMs issuing low-level command-line instructions, which can lead to incorrect implementations, Incalmo allows LLMs to specify high-level tasks (e.g., infect a host, scan a network), which are then carried out by Incalmo. Incalmo realizes these tasks by translating them into low-level primitives (e.g., commands to exploit tools). Incalmo also provides an environment state service and an attack graph service to provide structure to LLMs in selecting actions relevant to a multistage attack. Across 9 out of 10 realistic emulated networks (from 25 to 50 hosts), LLMs using Incalmo can successfully autonomously execute multistage attacks. We also conduct an ablation analysis to show the key role the high-level abstractions play. For instance, we find that both Incalmo's high-level tasks and services are crucial. Furthermore, even smaller-parameter LLMs with Incalmo can fully succeed in 5 of 10 environments, while larger-parameter LLMs without Incalmo do not fully succeed in any.
Large Language Models (LLMs) have been applied to various hardware design tasks, including Verilog code generation, EDA tool scripting, and RTL bug fixing. Despite this extensive exploration, LLMs are yet to be used for the task of post-synthesis metric reasoning and estimation of HDL designs. In this paper, we assess the ability of LLMs to reason about post-synthesis metrics of Verilog designs. We introduce MetRex, a large-scale dataset comprising 25,868 Verilog HDL designs and their corresponding post-synthesis metrics, namely area, delay, and static power. MetRex incorporates a Chain of Thought (CoT) template to enhance LLMs' reasoning about these metrics. Extensive experiments show that Supervised Fine-Tuning (SFT) boosts the LLM's reasoning capabilities on average by 37.0\%, 25.3\%, and 25.7\% on the area, delay, and static power, respectively. While SFT improves performance on our benchmark, it remains far from achieving optimal results, especially on complex problems. Comparing to state-of-the-art regression models, our approach delivers accurate post-synthesis predictions for 17.4\% more designs (within a 5\% error margin), in addition to offering a 1.7x speedup by eliminating the need for pre-processing. This work lays the groundwork for advancing LLM-based Verilog code metric reasoning.
In this paper, we assess the visualization literacy of two prominent Large Language Models (LLMs): OpenAI's Generative Pretrained Transformers (GPT), the backend of ChatGPT, and Google's Gemini, previously known as Bard, to establish benchmarks for assessing their visualization capabilities. While LLMs have shown promise in generating chart descriptions, captions, and design suggestions, their potential for evaluating visualizations remains under-explored. Collecting data from humans for evaluations has been a bottleneck for visualization research in terms of both time and money, and if LLMs were able to serve, even in some limited role, as evaluators, they could be a significant resource. To investigate the feasibility of using LLMs in the visualization evaluation process, we explore the extent to which LLMs possess visualization literacy -- a crucial factor for their effective utility in the field. We conducted a series of experiments using a modified 53-item Visualization Literacy Assessment Test (VLAT) for GPT-4 and Gemini. Our findings indicate that the LLMs we explored currently fail to achieve the same levels of visualization literacy when compared to data from the general public reported in VLAT, and LLMs heavily relied on their pre-existing knowledge to answer questions instead of utilizing the information provided by the visualization when answering questions.
The research in AI-based formal mathematical reasoning has shown an unstoppable growth trend. These studies have excelled in mathematical competitions like IMO, showing significant progress. However, these studies intertwined multiple skills simultaneously, i.e., problem-solving, reasoning, and writing formal specifications, making it hard to precisely identify the LLMs' strengths and weaknesses in each task. This paper focuses on formal verification, an immediate application scenario of formal reasoning, and decomposes it into six sub-tasks. We constructed 18k high-quality instruction-response pairs across five mainstream formal specification languages (Coq, Lean4, Dafny, ACSL, and TLA+) in six formal-verification-related tasks by distilling GPT-4o. They are split into a 14k+ fine-tuning dataset FM-alpaca and a 4k benchmark FM-Bench. We found that LLMs are good at writing proof segments when given either the code, or the detailed description of proof steps. Also, the fine-tuning brought about a nearly threefold improvement at most. Interestingly, we observed that fine-tuning with formal data also enhances mathematics, reasoning, and coding abilities. We hope our findings inspire further research. Fine-tuned models are released to facilitate subsequent studies
Nigeria is a multilingual country with 500+ languages. Naija is a Nigerian-Pidgin spoken by approx. 120M speakers in Nigeria and it is a mixed language (e.g., English, Portuguese, Yoruba, Hausa and Igbo). Although it has mainly been a spoken language until recently, there are now various platforms publishing exclusively in Naija such as Naija Wikipedia. However, it is hard to distinguish by non-native from a larger pidgin languages spoken across West Africa known as West African Pidgin English (WAPE) -- which is more simplied and understandable by wider audience in Ghana, Nigeria, and Cameroon. BBC news platform publishes exclusively in WAPE to cater for several countries in West Africa. In our paper, we show through statistical analyses and Machine Translation experiments that these two creole varieties do not represent each other (i.e., there are linguistic differences in word order and vocabulary) and Generative AI operates only based on WAPE. In other words, Naija is under-represented in Generative AI, and it is hard to teach LLMs with few examples.
Fixing Python dependency issues is a tedious and error-prone task for developers, who must manually identify and resolve environment dependencies and version constraints of third-party modules and Python interpreters. Researchers have attempted to automate this process by relying on large knowledge graphs and database lookup tables. However, these traditional approaches face limitations due to the variety of dependency error types, large sets of possible module versions, and conflicts among transitive dependencies. This study explores the potential of using large language models (LLMs) to automatically fix dependency issues in Python programs. We introduce PLLM (pronounced "plum"), a novel technique that employs retrieval-augmented generation (RAG) to help an LLM infer Python versions and required modules for a given Python file. PLLM builds a testing environment that iteratively (1) prompts the LLM for module combinations, (2) tests the suggested changes, and (3) provides feedback (error messages) to the LLM to refine the fix. This feedback cycle leverages natural language processing (NLP) to intelligently parse and interpret build error messages. We benchmark PLLM on the Gistable HG2.9K dataset, a collection of challenging single-file Python gists. We compare PLLM against two state-of-the-art automatic dependency inference approaches, namely PyEGo and ReadPyE, w.r.t. the ability to resolve dependency issues. Our results indicate that PLLM can fix more dependency issues than the two baselines, with +218 (+15.97%) more fixes over ReadPyE and +281 (+21.58%) over PyEGo. Our deeper analyses suggest that PLLM is particularly beneficial for projects with many dependencies and for specific third-party numerical and machine-learning modules. Our findings demonstrate the potential of LLM-based approaches to iteratively resolve Python dependency issues.
The widespread use of social media has led to a surge in popularity for automated methods of analyzing public opinion. Supervised methods are adept at text categorization, yet the dynamic nature of social media discussions poses a continual challenge for these techniques due to the constant shifting of the focus. On the other hand, traditional unsupervised methods for extracting themes from public discourse, such as topic modeling, often reveal overarching patterns that might not capture specific nuances. Consequently, a significant portion of research into social media discourse still depends on labor-intensive manual coding techniques and a human-in-the-loop approach, which are both time-consuming and costly. In this work, we study the problem of discovering arguments associated with a specific theme. We propose a generic LLMs-in-the-Loop strategy that leverages the advanced capabilities of Large Language Models (LLMs) to extract latent arguments from social media messaging. To demonstrate our approach, we apply our framework to contentious topics. We use two publicly available datasets: (1) the climate campaigns dataset of 14k Facebook ads with 25 themes and (2) the COVID-19 vaccine campaigns dataset of 9k Facebook ads with 14 themes. Additionally, we design a downstream task as stance prediction by leveraging talking points in climate debates. Furthermore, we analyze demographic targeting and the adaptation of messaging based on real-world events.
As autonomous agents become more prevalent, understanding their collective behaviour in strategic interactions is crucial. This study investigates the emergent cooperative tendencies of systems of Large Language Model (LLM) agents in a social dilemma. Unlike previous research where LLMs output individual actions, we prompt state-of-the-art LLMs to generate complete strategies for iterated Prisoner's Dilemma. Using evolutionary game theory, we simulate populations of agents with different strategic dispositions (aggressive, cooperative, or neutral) and observe their evolutionary dynamics. Our findings reveal that different LLMs exhibit distinct biases affecting the relative success of aggressive versus cooperative strategies. This research provides insights into the potential long-term behaviour of systems of deployed LLM-based autonomous agents and highlights the importance of carefully considering the strategic environments in which they operate.
Unit testing plays a pivotal role in the software development lifecycle, as it ensures code quality. However, writing high-quality unit tests remains a time-consuming task for developers in practice. More recently, the application of large language models (LLMs) in automated unit test generation has demonstrated promising results. Existing approaches primarily focus on interpreted programming languages (e.g., Java), while mature solutions tailored to compiled programming languages like C++ are yet to be explored. The intricate language features of C++, such as pointers, templates, and virtual functions, pose particular challenges for LLMs in generating both executable and high-coverage unit tests. To tackle the aforementioned problems, this paper introduces CITYWALK, a novel LLM-based framework for C++ unit test generation. CITYWALK enhances LLMs by providing a comprehensive understanding of the dependency relationships within the project under test via program analysis. Furthermore, CITYWALK incorporates language-specific knowledge about C++ derived from project documentation and empirical observations, significantly improving the correctness of the LLM-generated unit tests. We implement CITYWALK by employing the widely popular LLM GPT-4o. The experimental results show that CITYWALK outperforms current state-of-the-art approaches on a collection of eight popular C++ projects. Our findings demonstrate the effectiveness of CITYWALK in generating high-quality C++ unit tests.
In this paper, we present a system that uses a Large Language Model (LLM) to perform grammar and spelling correction as a component of Quality Assurance (QA) for texts generated by NLG systems, which is important for text production in real-world scenarios. Evaluating the results of the system on work-in-progress sports news texts in three languages, we show that it is able to deliver acceptable corrections.
Protecting online privacy requires users to engage with and comprehend website privacy policies, but many policies are difficult and tedious to read. We present PRISMe (Privacy Risk Information Scanner for Me), a novel Large Language Model (LLM)-driven privacy policy assessment tool, which helps users to understand the essence of a lengthy, complex privacy policy while browsing. The tool, a browser extension, integrates a dashboard and an LLM chat. One major contribution is the first rigorous evaluation of such a tool. In a mixed-methods user study (N=22), we evaluate PRISMe's efficiency, usability, understandability of the provided information, and impacts on awareness. While our tool improves privacy awareness by providing a comprehensible quick overview and a quality chat for in-depth discussion, users note issues with consistency and building trust in the tool. From our insights, we derive important design implications to guide future policy analysis tools.
Using large language models (LLMs) integration platforms without transparency about which LLM is being invoked can lead to potential security risks. Specifically, attackers may exploit this black-box scenario to deploy malicious models and embed viruses in the code provided to users. In this context, it is increasingly urgent for users to clearly identify the LLM they are interacting with, in order to avoid unknowingly becoming victims of malicious models. However, existing studies primarily focus on mixed classification of human and machine-generated text, with limited attention to classifying texts generated solely by different models. Current research also faces dual bottlenecks: poor quality of LLM-generated text (LLMGT) datasets and limited coverage of detectable LLMs, resulting in poor detection performance for various LLMGT in black-box scenarios. We propose the first LLMGT fingerprint detection model, \textbf{FDLLM}, based on Qwen2.5-7B and fine-tuned using LoRA to address these challenges. FDLLM can more efficiently handle detection tasks across multilingual and multi-domain scenarios. Furthermore, we constructed a dataset named \textbf{FD-Datasets}, consisting of 90,000 samples that span multiple languages and domains, covering 20 different LLMs. Experimental results demonstrate that FDLLM achieves a macro F1 score 16.7\% higher than the best baseline method, LM-D.
Current Large Language Models (LLMs) cannot support users to precisely balance privacy protection and output performance during individual consultations. We introduce Adanonymizer, an anonymization plug-in that allows users to control this balance by navigating a trade-off curve. A survey (N=221) revealed a privacy paradox, where users frequently disclosed sensitive information despite acknowledging privacy risks. The study further demonstrated that privacy risks were not significantly correlated with model output performance, highlighting the potential to navigate this trade-off. Adanonymizer normalizes privacy and utility ratings by type and automates the pseudonymization of sensitive terms based on user preferences, significantly reducing user effort. Its 2D color palette interface visualizes the privacy-utility trade-off, allowing users to adjust the balance by manipulating a point. An evaluation (N=36) compared Adanonymizer with ablation methods and differential privacy techniques, where Adanonymizer significantly reduced modification time, achieved better perceived model performance and overall user preference.
Large language models (LLMs) have shown remarkable potential as autonomous agents, particularly in web-based tasks. However, existing LLM web agents heavily rely on expensive proprietary LLM APIs, while open LLMs lack the necessary decision-making capabilities. This paper introduces WebRL, a self-evolving online curriculum reinforcement learning framework designed to train high-performance web agents using open LLMs. WebRL addresses three key challenges in building LLM web agents, including the scarcity of training tasks, sparse feedback signals, and policy distribution drift in online learning. Specifically, WebRL incorporates 1) a self-evolving curriculum that generates new tasks from unsuccessful attempts, 2) a robust outcome-supervised reward model (ORM), and 3) adaptive reinforcement learning strategies to ensure consistent improvements. We apply WebRL to transform open Llama-3.1 and GLM-4 models into proficient web agents. On WebArena-Lite, WebRL improves the success rate of Llama-3.1-8B from 4.8% to 42.4%, and from 6.1% to 43% for GLM-4-9B. These open models significantly surpass the performance of GPT-4-Turbo (17.6%) and GPT-4o (13.9%) and outperform previous state-of-the-art web agents trained on open LLMs (AutoWebGLM, 18.2%). Our findings demonstrate WebRL's effectiveness in bridging the gap between open and proprietary LLM-based web agents, paving the way for more accessible and powerful autonomous web interaction systems.
The release of ChatGPT in November 2022 sparked an explosion of interest in post-training and an avalanche of new preference optimization (PO) methods. These methods claim superior alignment by virtue of better correspondence with human pairwise preferences, often measured by LLM-judges. In this work, we attempt to answer the following question -- do LLM-judge preferences translate to progress on other, more concrete metrics for alignment, and if not, why not? We define a concrete metric for alignment, and introduce SOS-Bench (Substance Outweighs Style Benchmark), which is to the best of our knowledge the largest standardized, reproducible LLM meta-benchmark to date. We find that (1) LLM-judge preferences do not correlate with concrete measures of safety, world knowledge, and instruction following; (2) LLM-judges have powerful implicit biases, prioritizing style over factuality and safety; and (3) the supervised fine-tuning (SFT) stage of post-training, and not the PO stage, has the greatest impact on alignment, with data scaling and prompt diversity as the driving factors. Our codebase and complete results can be found at https://github.com/penfever/sos-bench.
Deploying large language models (LLMs) of several billion parameters can be impractical in most industrial use cases due to constraints such as cost, latency limitations, and hardware accessibility. Knowledge distillation (KD) offers a solution by compressing knowledge from resource-intensive large models to smaller ones. Various strategies exist, some relying on the text generated by the teacher model and optionally utilizing his logits to enhance learning. However, these methods based on logits often require both teacher and student models to share the same tokenizer, limiting their applicability across different LLM families. In this paper, we introduce Universal Logit Distillation (ULD) loss, grounded in optimal transport, to address this limitation. Our experimental results demonstrate the effectiveness of ULD loss in enabling distillation across models with different architectures and tokenizers, paving the way to a more widespread use of distillation techniques.
Mobile robot path planning in complex environments remains a significant challenge, especially in achieving efficient, safe and robust paths. The traditional path planning techniques like DRL models typically trained for a given configuration of the starting point and target positions, these models only perform well when these conditions are satisfied. In this paper, we proposed a novel path planning framework that embeds Large Language Models to empower mobile robots with the capability of dynamically interpreting natural language commands and autonomously generating efficient, collision-free navigation paths. The proposed framework uses LLMs to translate high-level user inputs into actionable waypoints while dynamically adjusting paths in response to obstacles. We experimentally evaluated our proposed LLM-based approach across three different environments of progressive complexity, showing the robustness of our approach with llama3.1 model that outperformed other LLM models in path planning time, waypoint generation success rate, and collision avoidance. This underlines the promising contribution of LLMs for enhancing the capability of mobile robots, especially when their operation involves complex decisions in large and complex environments. Our framework has provided safer, more reliable navigation systems and opened a new direction for the future research. The source code of this work is publicly available on GitHub.
The rise of large language models (LLMs) has led to more diverse and higher-quality machine-generated text. However, their high expressive power makes it difficult to control outputs based on specific business instructions. In response, benchmarks focusing on the controllability of LLMs have been developed, but several issues remain: (1) They primarily cover major languages like English and Chinese, neglecting low-resource languages like Japanese; (2) Current benchmarks employ task-specific evaluation metrics, lacking a unified framework for selecting models based on controllability across different use cases. To address these challenges, this research introduces LCTG Bench, the first Japanese benchmark for evaluating the controllability of LLMs. LCTG Bench provides a unified framework for assessing control performance, enabling users to select the most suitable model for their use cases based on controllability. By evaluating nine diverse Japanese-specific and multilingual LLMs like GPT-4, we highlight the current state and challenges of controllability in Japanese LLMs and reveal the significant gap between multilingual models and Japanese-specific models.
Broad adoption of Large Language Models (LLM) demands rapid expansions of cloud LLM inference clusters, leading to accumulation of embodied carbon$-$the emissions from manufacturing and supplying IT assets$-$that mostly concentrate on inference server CPU. This paper delves into the challenges of sustainable growth of cloud LLM inference, emphasizing extended amortization of CPU embodied over an increased lifespan. Given the reliability risks of silicon aging, we propose an aging-aware CPU core management technique to delay CPU aging effects, allowing the cluster operator to safely increase CPU life. Our technique exploits CPU underutilization patterns that we uncover in cloud LLM inference by halting aging in unused cores and even-outing aging in active cores via selective deep idling and aging-aware inference task allocation. Through extensive simulations using real-world Azure inference traces and an extended LLM cluster simulator from Microsoft, we show superior performance of our technique over existing methods with an estimated 37.67\% reduction in yearly embodied carbon emissions through p99 performance of managing CPU aging effects, a 77\% reduction in CPU underutilization, and less than 10\% impact to the inference service quality.
The effectiveness of automatic evaluation of generative models is typically measured by comparing the labels generated via automation with labels by humans using correlation metrics. However, metrics like Krippendorff's $\alpha$ and Randolph's $\kappa$ were originally designed to measure the reliability of human labeling, thus make assumptions about typical human labeling behavior, and these assumptions may not be applicable to machine generated labels. In this paper, we show how *relying on a single aggregate correlation score* can obscure fundamental differences between human labels and those from automatic evaluation, including LLM-as-a-Judge. Specifically, we demonstrate that when the proportion of samples with variation or uncertainty in human assigned labels is relatively high, machine labels (generated by automatic evaluation methods) may superficially appear to have similar or better correlation with the human majority label compared to the human-to-human (HH) correlation. This can create the illusion that labels from automatic evaluation approximates the human majority label. However, as the proportion of samples with consistent human labels increases, the correlation between machine and human labels fall well below HH correlation. Based on these findings, we first propose stratifying data by human label uncertainty to provide a more robust analysis of automatic evaluation performance. Second, recognizing that uncertainty and variation are inherent in perception-based human evaluations, such as those involving attitudes or preferences, we introduce a new metric - binned Jensen-Shannon Divergence for perception for such scenarios to better measure the effectiveness of automatic evaluations. We present visualization techniques -- perception charts, to contextualize correlation measures appropriately. We have open-sourced at https://github.com/amazon-science/BeyondCorrelation.
Recent studies developed jailbreaking attacks, which construct jailbreaking prompts to fool LLMs into responding to harmful questions. Early-stage jailbreaking attacks require access to model internals or significant human efforts. More advanced attacks utilize genetic algorithms for automatic and black-box attacks. However, the random nature of genetic algorithms significantly limits the effectiveness of these attacks. In this paper, we propose RLbreaker, a black-box jailbreaking attack driven by deep reinforcement learning (DRL). We model jailbreaking as a search problem and design an RL agent to guide the search, which is more effective and has less randomness than stochastic search, such as genetic algorithms. Specifically, we design a customized DRL system for the jailbreaking problem, including a novel reward function and a customized proximal policy optimization (PPO) algorithm. Through extensive experiments, we demonstrate that RLbreaker is much more effective than existing jailbreaking attacks against six state-of-the-art (SOTA) LLMs. We also show that RLbreaker is robust against three SOTA defenses and its trained agents can transfer across different LLMs. We further validate the key design choices of RLbreaker via a comprehensive ablation study.
Large Language Models (LLMs) have demonstrated remarkable performance across diverse tasks. LLMs continue to be vulnerable to external threats, particularly Denial-of-Service (DoS) attacks. Specifically, LLM-DoS attacks aim to exhaust computational resources and block services. However, prior works tend to focus on performing white-box attacks, overlooking black-box settings. In this work, we propose an automated algorithm designed for black-box LLMs, called Auto-Generation for LLM-DoS Attack (AutoDoS). AutoDoS introduces DoS Attack Tree and optimizes the prompt node coverage to enhance effectiveness under black-box conditions. Our method can bypass existing defense with enhanced stealthiness via semantic improvement of prompt nodes. Furthermore, we reveal that implanting Length Trojan in Basic DoS Prompt aids in achieving higher attack efficacy. Experimental results show that AutoDoS amplifies service response latency by over 250 $\times \uparrow$, leading to severe resource consumption in terms of GPU utilization and memory usage. Our code is available at https://github.com/shuita2333/AutoDoS.
The growing importance of textual and relational systems has driven interest in enhancing large language models (LLMs) for graph-structured data, particularly Text-Attributed Graphs (TAGs), where samples are represented by textual descriptions interconnected by edges. While research has largely focused on developing specialized graph LLMs through task-specific instruction tuning, a comprehensive benchmark for evaluating LLMs solely through prompt design remains surprisingly absent. Without such a carefully crafted evaluation benchmark, most if not all, tailored graph LLMs are compared against general LLMs using simplistic queries (e.g., zero-shot reasoning with LLaMA), which can potentially camouflage many advantages as well as unexpected predicaments of them. To achieve more general evaluations and unveil the true potential of LLMs for graph tasks, we introduce Graph In-context Learning (GraphICL) Benchmark, a comprehensive benchmark comprising novel prompt templates designed to capture graph structure and handle limited label knowledge. Our systematic evaluation shows that general-purpose LLMs equipped with our GraphICL outperform state-of-the-art specialized graph LLMs and graph neural network models in resource-constrained settings and out-of-domain tasks. These findings highlight the significant potential of prompt engineering to enhance LLM performance on graph learning tasks without training and offer a strong baseline for advancing research in graph LLMs.
Intelligent Tutoring Systems (ITSs) have revolutionized education by offering personalized learning experiences. However, as goal-oriented learning, which emphasizes efficiently achieving specific objectives, becomes increasingly important in professional contexts, existing ITSs often struggle to deliver this type of targeted learning experience. In this paper, we propose GenMentor, an LLM-powered multi-agent framework designed to deliver goal-oriented, personalized learning within ITS. GenMentor begins by accurately mapping learners' goals to required skills using a fine-tuned LLM trained on a custom goal-to-skill dataset. After identifying the skill gap, it schedules an efficient learning path using an evolving optimization approach, driven by a comprehensive and dynamic profile of learners' multifaceted status. Additionally, GenMentor tailors learning content with an exploration-drafting-integration mechanism to align with individual learner needs. Extensive automated and human evaluations demonstrate GenMentor's effectiveness in learning guidance and content quality. Furthermore, we have deployed it in practice and also implemented it as an application. Practical human study with professional learners further highlights its effectiveness in goal alignment and resource targeting, leading to enhanced personalization. Supplementary resources are available at https://github.com/GeminiLight/gen-mentor.
We present a novel class of jailbreak adversarial attacks on LLMs, termed Task-in-Prompt (TIP) attacks. Our approach embeds sequence-to-sequence tasks (e.g., cipher decoding, riddles, code execution) into the model's prompt to indirectly generate prohibited inputs. To systematically assess the effectiveness of these attacks, we introduce the PHRYGE benchmark. We demonstrate that our techniques successfully circumvent safeguards in six state-of-the-art language models, including GPT-4o and LLaMA 3.2. Our findings highlight critical weaknesses in current LLM safety alignments and underscore the urgent need for more sophisticated defence strategies. Warning: this paper contains examples of unethical inquiries used solely for research purposes.
The reasoning abilities of Large Language Models (LLMs) can be improved by structurally denoising their weights, yet existing techniques primarily focus on denoising the feed-forward network (FFN) of the transformer block, and can not efficiently utilise the Multi-head Attention (MHA) block, which is the core of transformer architectures. To address this issue, we propose a novel intuitive framework that, at its very core, performs MHA compression through a multi-head tensorisation process and the Tucker decomposition. This enables both higher-dimensional structured denoising and compression of the MHA weights, by enforcing a shared higher-dimensional subspace across the weights of the multiple attention heads. We demonstrate that this approach consistently enhances the reasoning capabilities of LLMs across multiple benchmark datasets, and for both encoder-only and decoder-only architectures, while achieving compression rates of up to $\sim 250$ times in the MHA weights, all without requiring any additional data, training, or fine-tuning. Furthermore, we show that the proposed method can be seamlessly combined with existing FFN-only-based denoising techniques to achieve further improvements in LLM reasoning performance.
Large language models (LLMs) are increasingly utilized for machine translation, yet their predictions often exhibit uncertainties that hinder interpretability and user trust. Effectively visualizing these uncertainties can enhance the usability of LLM outputs, particularly in contexts where translation accuracy is critical. This paper addresses two primary objectives: (1) providing users with token-level insights into model confidence and (2) developing a web-based visualization tool to quantify and represent translation uncertainties. To achieve these goals, we utilized the T5 model with the WMT19 dataset for translation tasks and evaluated translation quality using established metrics such as BLEU, METEOR, and ROUGE. We introduced three novel uncertainty quantification (UQ) metrics: (1) the geometric mean of token probabilities, (2) the arithmetic mean of token probabilities, and (3) the arithmetic mean of the kurtosis of token distributions. These metrics provide a simple yet effective framework for evaluating translation performance. Our analysis revealed a linear relationship between the traditional evaluation metrics and our UQ metrics, demonstrating the validity of our approach. Additionally, we developed an interactive web-based visualization that uses a color gradient to represent token confidence. This tool offers users a clear and intuitive understanding of translation quality while providing valuable insights into model performance. Overall, we show that our UQ metrics and visualization are both robust and interpretable, offering practical tools for evaluating and accessing machine translation systems.
Large Language Models (LLMs) have excelled in various tasks but are still vulnerable to jailbreaking attacks, where attackers create jailbreak prompts to mislead the model to produce harmful or offensive content. Current jailbreak methods either rely heavily on manually crafted templates, which pose challenges in scalability and adaptability, or struggle to generate semantically coherent prompts, making them easy to detect. Additionally, most existing approaches involve lengthy prompts, leading to higher query costs.In this paper, to remedy these challenges, we introduce a novel jailbreaking attack framework called PAPILLON, which is an automated, black-box jailbreaking attack framework that adapts the black-box fuzz testing approach with a series of customized designs. Instead of relying on manually crafted templates,PAPILLON starts with an empty seed pool, removing the need to search for any related jailbreaking templates. We also develop three novel question-dependent mutation strategies using an LLM helper to generate prompts that maintain semantic coherence while significantly reducing their length. Additionally, we implement a two-level judge module to accurately detect genuine successful jailbreaks. We evaluated PAPILLON on 7 representative LLMs and compared it with 5 state-of-the-art jailbreaking attack strategies. For proprietary LLM APIs, such as GPT-3.5 turbo, GPT-4, and Gemini-Pro, PAPILLONs achieves attack success rates of over 90%, 80%, and 74%, respectively, exceeding existing baselines by more than 60\%. Additionally, PAPILLON can maintain high semantic coherence while significantly reducing the length of jailbreak prompts. When targeting GPT-4, PAPILLON can achieve over 78% attack success rate even with 100 tokens. Moreover, PAPILLON demonstrates transferability and is robust to state-of-the-art defenses.
Recent advancements in large language models (LLMs) promise to expand mental health interventions by emulating therapeutic techniques, potentially easing barriers to care. Yet there is a lack of real-world empirical evidence evaluating the strengths and limitations of LLM-enabled psychotherapy interventions. In this work, we evaluate an LLM-powered chatbot, designed via prompt engineering to deliver cognitive restructuring (CR), with 19 users. Mental health professionals then examined the resulting conversation logs to uncover potential benefits and pitfalls. Our findings indicate that an LLM-based CR approach has the capability to adhere to core CR protocols, prompt Socratic questioning, and provide empathetic validation. However, issues of power imbalances, advice-giving, misunderstood cues, and excessive positivity reveal deeper challenges, including the potential to erode therapeutic rapport and ethical concerns. We also discuss design implications for leveraging LLMs in psychotherapy and underscore the importance of expert oversight to mitigate these concerns, which are critical steps toward safer, more effective AI-assisted interventions.
Large Language Models (LLMs) have shown promising results in a variety of literary tasks, often using complex memorized details of narration and fictional characters. In this work, we evaluate the ability of Llama-3 at attributing utterances of direct-speech to their speaker in novels. The LLM shows impressive results on a corpus of 28 novels, surpassing published results with ChatGPT and encoder-based baselines by a large margin. We then validate these results by assessing the impact of book memorization and annotation contamination. We found that these types of memorization do not explain the large performance gain, making Llama-3 the new state-of-the-art for quotation attribution in English literature. We release publicly our code and data.
Despite recent advances in multimodal large language models (MLLMs), their development has predominantly focused on English- and western-centric datasets and tasks, leaving most of the world's languages and diverse cultural contexts underrepresented. This paper introduces Pangea, a multilingual multimodal LLM trained on PangeaIns, a diverse 6M instruction dataset spanning 39 languages. PangeaIns features: 1) high-quality English instructions, 2) carefully machine-translated instructions, and 3) culturally relevant multimodal tasks to ensure cross-cultural coverage. To rigorously assess models' capabilities, we introduce PangeaBench, a holistic evaluation suite encompassing 14 datasets covering 47 languages. Results show that Pangea significantly outperforms existing open-source models in multilingual settings and diverse cultural contexts. Ablation studies further reveal the importance of English data proportions, language popularity, and the number of multimodal training samples on overall performance. We fully open-source our data, code, and trained checkpoints, to facilitate the development of inclusive and robust multilingual MLLMs, promoting equity and accessibility across a broader linguistic and cultural spectrum.
Large Language Models (LLMs) have demonstrated great potential in complex reasoning tasks, yet they fall short when tackling more sophisticated challenges, especially when interacting with environments through generating executable actions. This inadequacy primarily stems from the lack of built-in action knowledge in language agents, which fails to effectively guide the planning trajectories during task solving and results in planning hallucination. To address this issue, we introduce KnowAgent, a novel approach designed to enhance the planning capabilities of LLMs by incorporating explicit action knowledge. Specifically, KnowAgent employs an action knowledge base and a knowledgeable self-learning strategy to constrain the action path during planning, enabling more reasonable trajectory synthesis, and thereby enhancing the planning performance of language agents. Experimental results on HotpotQA and ALFWorld based on various backbone models demonstrate that KnowAgent can achieve comparable or superior performance to existing baselines. Further analysis indicates the effectiveness of KnowAgent in terms of planning hallucinations mitigation. Code is available in https://github.com/zjunlp/KnowAgent.
Classroom debates are a unique form of collaborative learning characterized by fast-paced, high-intensity interactions that foster critical thinking and teamwork. Despite the recognized importance of debates, the role of AI tools, particularly LLM-based systems, in supporting this dynamic learning environment has been under-explored in HCI. This study addresses this opportunity by investigating the integration of LLM-based AI into real-time classroom debates. Over four weeks, 22 students in a Design History course participated in three rounds of debates with support from ChatGPT. The findings reveal how learners prompted the AI to offer insights, collaboratively processed its outputs, and divided labor in team-AI interactions. The study also surfaces key advantages of AI usage, reducing social anxiety, breaking communication barriers, and providing scaffolding for novices, alongside risks, such as information overload and cognitive dependency, which could limit learners' autonomy. We thereby discuss a set of nuanced implications for future HCI exploration.
Existing research primarily evaluates the values of LLMs by examining their stated inclinations towards specific values. However, the "Value-Action Gap," a phenomenon rooted in environmental and social psychology, reveals discrepancies between individuals' stated values and their actions in real-world contexts. To what extent do LLMs exhibit a similar gap between their stated values and their actions informed by those values? This study introduces ValueActionLens, an evaluation framework to assess the alignment between LLMs' stated values and their value-informed actions. The framework encompasses the generation of a dataset comprising 14.8k value-informed actions across twelve cultures and eleven social topics, and two tasks to evaluate how well LLMs' stated value inclinations and value-informed actions align across three different alignment measures. Extensive experiments reveal that the alignment between LLMs' stated values and actions is sub-optimal, varying significantly across scenarios and models. Analysis of misaligned results identifies potential harms from certain value-action gaps. To predict the value-action gaps, we also uncover that leveraging reasoned explanations improves performance. These findings underscore the risks of relying solely on the LLMs' stated values to predict their behaviors and emphasize the importance of context-aware evaluations of LLM values and value-action gaps.
Customizable role-playing in large language models (LLMs), also known as character generalization, is gaining increasing attention for its versatility and cost-efficiency in developing and deploying role-playing dialogue agents. This study explores a large-scale data synthesis approach to equip LLMs with character generalization capabilities. We begin by synthesizing large-scale character profiles using personas from Persona Hub and then explore two strategies: response rewriting and response generation, to create character-aligned instructional responses. To validate the effectiveness of our synthetic instruction tuning data for character generalization, we perform supervised fine-tuning (SFT) using the LLaMA-3 8B model. Our best-performing model strengthens the original LLaMA-3 8B Instruct model and achieves performance comparable to GPT-4o models on role-playing dialogue. We release our synthetic characters and instruction-tuning dialogues to support public research.
Ontology matching (OM) enables semantic interoperability between different ontologies and resolves their conceptual heterogeneity by aligning related entities. OM systems currently have two prevailing design paradigms: conventional knowledge-based expert systems and newer machine learning-based predictive systems. While large language models (LLMs) and LLM agents have revolutionised data engineering and have been applied creatively in many domains, their potential for OM remains underexplored. This study introduces a novel agent-powered LLM-based design paradigm for OM systems. With consideration of several specific challenges in leveraging LLM agents for OM, we propose a generic framework, namely Agent-OM (Agent for Ontology Matching), consisting of two Siamese agents for retrieval and matching, with a set of OM tools. Our framework is implemented in a proof-of-concept system. Evaluations of three Ontology Alignment Evaluation Initiative (OAEI) tracks over state-of-the-art OM systems show that our system can achieve results very close to the long-standing best performance on simple OM tasks and can significantly improve the performance on complex and few-shot OM tasks.
Various benchmarks have been proposed to assess the performance of large language models (LLMs) in different coding scenarios. We refer to them as code-related benchmarks. However, there are no systematic guidelines by which such a benchmark should be developed to ensure its quality, reliability, and reproducibility. We propose How2Bench, which is comprised of a 55- 55-criteria checklist as a set of guidelines to govern the development of code-related benchmarks comprehensively. Using HOW2BENCH, we profiled 274 benchmarks released within the past decade and found concerning issues. Nearly 70% of the benchmarks did not take measures for data quality assurance; over 10% did not even open source or only partially open source. Many highly cited benchmarks have loopholes, including duplicated samples, incorrect reference codes/tests/prompts, and unremoved sensitive/confidential information. Finally, we conducted a human study involving 49 participants, which revealed significant gaps in awareness of the importance of data quality, reproducibility, and transparency.
Large Language Models (LLMs) excel in text generation and understanding, especially in simulating socio-political and economic patterns, serving as an alternative to traditional surveys. However, their global applicability remains questionable due to unexplored biases across socio-demographic and geographic contexts. This study examines how LLMs perform across diverse populations by analyzing public surveys from Chile and the United States, focusing on predictive accuracy and fairness metrics. The results show performance disparities, with LLM consistently outperforming on U.S. datasets. This bias originates from the U.S.-centric training data, remaining evident after accounting for socio-demographic differences. In the U.S., political identity and race significantly influence prediction accuracy, while in Chile, gender, education, and religious affiliation play more pronounced roles. Our study presents a novel framework for measuring socio-demographic biases in LLMs, offering a path toward ensuring fairer and more equitable model performance across diverse socio-cultural contexts.
We propose SLoPe, a Double-Pruned Sparse Plus Lazy Low-rank Adapter Pretraining method for LLMs that improves the accuracy of sparse LLMs while accelerating their pretraining and inference and reducing their memory footprint. Sparse pretraining of LLMs reduces the accuracy of the model, to overcome this, prior work uses dense models during fine-tuning. SLoPe improves the accuracy of sparsely pretrained models by adding low-rank adapters in the final 1% iterations of pretraining without adding significant overheads to the model pretraining and inference. In addition, SLoPe uses a double-pruned backward pass formulation that prunes the transposed weight matrix using N:M sparsity structures to enable an accelerated sparse backward pass. SLoPe accelerates the training and inference of models with billions of parameters up to $1.25\times$ and $1.54\times$ respectively (OPT-33B and OPT-66B) while reducing their memory usage by up to $0.63\times$ and $0.61\times$ for training and inference respectively.
The global adoption of Large Language Models (LLMs) in healthcare shows promise to enhance clinical workflows and improve patient outcomes. However, Automatic Speech Recognition (ASR) errors in critical medical terms remain a significant challenge. These errors can compromise patient care and safety if not detected. This study investigates the prevalence and impact of ASR errors in medical transcription in Nigeria, the United Kingdom, and the United States. By evaluating raw and LLM-corrected transcriptions of accented English in these regions, we assess the potential and limitations of LLMs to address challenges related to accents and medical terminology in ASR. Our findings highlight significant disparities in ASR accuracy across regions and identify specific conditions under which LLM corrections are most effective.
Developing effective text summarizers remains a challenge due to issues like hallucinations, key information omissions, and verbosity in LLM-generated summaries. This work explores using LLM-generated feedback to improve summary quality by aligning the summaries with human preferences for faithfulness, completeness, and conciseness. We introduce FeedSum, a large-scale dataset containing multi-dimensional LLM feedback on summaries of varying quality across diverse domains. Our experiments show how feedback quality, dimensionality, and granularity influence preference learning, revealing that high-quality, multi-dimensional, fine-grained feedback significantly improves summary generation. We also compare two methods for using this feedback: supervised fine-tuning and direct preference optimization. Finally, we introduce SummLlama3-8b, a model that outperforms the nearly 10x larger Llama3-70b-instruct in generating human-preferred summaries, demonstrating that smaller models can achieve superior performance with appropriate training. The full dataset and SummLlama3-8B model are available at https://huggingface.co/datasets/DISLab/FeedSum and https://huggingface.co/DISLab/SummLlama3-8B.
Considering the significance of proteins, computational protein science has always been a critical scientific field, dedicated to revealing knowledge and developing applications within the protein sequence-structure-function paradigm. In the last few decades, Artificial Intelligence (AI) has made significant impacts in computational protein science, leading to notable successes in specific protein modeling tasks. However, those previous AI models still meet limitations, such as the difficulty in comprehending the semantics of protein sequences, and the inability to generalize across a wide range of protein modeling tasks. Recently, LLMs have emerged as a milestone in AI due to their unprecedented language processing & generalization capability. They can promote comprehensive progress in fields rather than solving individual tasks. As a result, researchers have actively introduced LLM techniques in computational protein science, developing protein Language Models (pLMs) that skillfully grasp the foundational knowledge of proteins and can be effectively generalized to solve a diversity of sequence-structure-function reasoning problems. While witnessing prosperous developments, it's necessary to present a systematic overview of computational protein science empowered by LLM techniques. First, we summarize existing pLMs into categories based on their mastered protein knowledge, i.e., underlying sequence patterns, explicit structural and functional information, and external scientific languages. Second, we introduce the utilization and adaptation of pLMs, highlighting their remarkable achievements in promoting protein structure prediction, protein function prediction, and protein design studies. Then, we describe the practical application of pLMs in antibody design, enzyme design, and drug discovery. Finally, we specifically discuss the promising future directions in this fast-growing field.
Aerospace manufacturing demands exceptionally high precision in technical parameters. The remarkable performance of Large Language Models (LLMs), such as GPT-4 and QWen, in Natural Language Processing has sparked industry interest in their application to tasks including process design, material selection, and tool information retrieval. However, LLMs are prone to generating "hallucinations" in specialized domains, producing inaccurate or false information that poses significant risks to the quality of aerospace products and flight safety. This paper introduces a set of evaluation metrics tailored for LLMs in aerospace manufacturing, aiming to assess their accuracy by analyzing their performance in answering questions grounded in professional knowledge. Firstly, key information is extracted through in-depth textual analysis of classic aerospace manufacturing textbooks and guidelines. Subsequently, utilizing LLM generation techniques, we meticulously construct multiple-choice questions with multiple correct answers of varying difficulty. Following this, different LLM models are employed to answer these questions, and their accuracy is recorded. Experimental results demonstrate that the capabilities of LLMs in aerospace professional knowledge are in urgent need of improvement. This study provides a theoretical foundation and practical guidance for the application of LLMs in aerospace manufacturing, addressing a critical gap in the field.
Code review remains a critical yet resource-intensive process in software development, particularly challenging in large-scale industrial environments. While Large Language Models (LLMs) show promise for automating code review, existing solutions face significant limitations in precision and practicality. This paper presents BitsAI-CR, an innovative framework that enhances code review through a two-stage approach combining RuleChecker for initial issue detection and ReviewFilter for precision verification. The system is built upon a comprehensive taxonomy of review rules and implements a data flywheel mechanism that enables continuous performance improvement through structured feedback and evaluation metrics. Our approach introduces an Outdated Rate metric that can reflect developers' actual adoption of review comments, enabling automated evaluation and systematic optimization at scale. Empirical evaluation demonstrates BitsAI-CR's effectiveness, achieving 75.0% precision in review comment generation. For the Go language which has predominant usage at ByteDance, we maintain an Outdated Rate of 26.7%. The system has been successfully deployed at ByteDance, serving over 12,000 Weekly Active Users (WAU). Our work provides valuable insights into the practical application of automated code review and offers a blueprint for organizations seeking to implement automated code reviews at scale.
The rapid advancement of large language models (LLMs) in biological-medical applications has highlighted a gap between their potential and the limited scale and often low quality of available open-source annotated textual datasets. In addition, the inherent complexity of the biomedical knowledge hierarchy significantly hampers efforts to bridge this gap.Can LLMs themselves play a pivotal role in overcoming this limitation? Motivated by this question, we investigate this challenge in the present study.We propose a framework that automates the distillation of high-quality textual training data from the extensive scientific literature. Our approach self-evaluates and generates questions that are more closely aligned with the biomedical domain, guided by the biomedical knowledge hierarchy through medical subject headings (MeSH). This comprehensive framework establishes an automated workflow, thereby eliminating the need for manual intervention. Furthermore, we conducted comprehensive experiments to evaluate the impact of our framework-generated data on downstream language models of varying sizes. Our approach substantially improves question-answering tasks compared to pre-trained models from the life sciences domain and powerful close-source models represented by GPT-4. Notably, the generated AI-Ready dataset enabled the Llama3-70B base model to outperform GPT-4 using MedPrompt with multiple times the number of parameters. Detailed case studies and ablation experiments underscore the significance of each component within our framework
Large Language Models for sequential recommendation (LLM4SR), which transform user-item interactions into language modeling, have shown promising results. However, due to the limitations of context window size and the computational costs associated with Large Language Models (LLMs), current approaches primarily truncate user history by only considering the textual information of items from the most recent interactions in the input prompt. This truncation fails to fully capture the long-term behavioral patterns of users. To address this, we propose a multi-grained patching framework -- PatchRec. It compresses the textual tokens of an item title into a compact item patch, and further compresses multiple item patches into a denser session patch, with earlier interactions being compressed to a greater degree. The framework consists of two stages: (1) Patch Pre-training, which familiarizes LLMs with item-level compression patterns, and (2) Patch Fine-tuning, which teaches LLMs to model sequences at multiple granularities. Through this simple yet effective approach, empirical results demonstrate that PatchRec outperforms existing methods, achieving significant performance gains with fewer tokens fed to the LLM. Specifically, PatchRec shows up to a 32% improvement in HR@20 on the Goodreads dataset over uncompressed baseline, while using only 7% of the tokens. This multi-grained sequence modeling paradigm, with an adjustable compression ratio, enables LLMs to be efficiently deployed in real-world recommendation systems that handle extremely long user behavior sequences.
Research question answering requires accurate retrieval and contextual understanding of scientific literature. However, current Retrieval-Augmented Generation (RAG) methods often struggle to balance complex document relationships with precise information retrieval. In this paper, we introduce Contextualized Graph Retrieval-Augmented Generation (CG-RAG), a novel framework that integrates sparse and dense retrieval signals within graph structures to enhance retrieval efficiency and subsequently improve generation quality for research question answering. First, we propose a contextual graph representation for citation graphs, effectively capturing both explicit and implicit connections within and across documents. Next, we introduce Lexical-Semantic Graph Retrieval (LeSeGR), which seamlessly integrates sparse and dense retrieval signals with graph encoding. It bridges the gap between lexical precision and semantic understanding in citation graph retrieval, demonstrating generalizability to existing graph retrieval and hybrid retrieval methods. Finally, we present a context-aware generation strategy that utilizes the retrieved graph-structured information to generate precise and contextually enriched responses using large language models (LLMs). Extensive experiments on research question answering benchmarks across multiple domains demonstrate that our CG-RAG framework significantly outperforms RAG methods combined with various state-of-the-art retrieval approaches, delivering superior retrieval accuracy and generation quality.
In recent years, large language models (LLMs) have seen rapid advancements, significantly impacting various fields such as computer vision, natural language processing, and software engineering. These LLMs, exemplified by OpenAI's ChatGPT, have revolutionized the way we approach language understanding and generation tasks. However, in contrast to traditional software development practices, LLM development introduces new challenges for AI developers in design, implementation, and deployment. These challenges span different areas (such as prompts, APIs, and plugins), requiring developers to navigate unique methodologies and considerations specific to LLM application development. Despite the profound influence of LLMs, to the best of our knowledge, these challenges have not been thoroughly investigated in previous empirical studies. To fill this gap, we present the first comprehensive study on understanding the challenges faced by LLM developers. Specifically, we crawl and analyze 29,057 relevant questions from a popular OpenAI developer forum. We first examine their popularity and difficulty. After manually analyzing 2,364 sampled questions, we construct a taxonomy of challenges faced by LLM developers. Based on this taxonomy, we summarize a set of findings and actionable implications for LLM-related stakeholders, including developers and providers (especially the OpenAI organization).
Key-Value (KV) cache facilitates efficient large language models (LLMs) inference by avoiding recomputation of past KVs. As the batch size and context length increase, the oversized KV caches become a significant memory bottleneck, highlighting the need for efficient compression. Existing KV quantization rely on fine-grained quantization or the retention of a significant portion of high bit-widths caches, both of which compromise compression ratio and often fail to maintain robustness at extremely low average bit-widths. In this work, we explore the potential of rotation technique for 2-bit KV quantization and propose RotateKV, which achieves accurate and robust performance through the following innovations: (i) Outlier-Aware Rotation, which utilizes channel-reordering to adapt the rotations to varying channel-wise outlier distributions without sacrificing the computational efficiency of the fast Walsh-Hadamard transform (FWHT); (ii) Pre-RoPE Grouped-Head Rotation, which mitigates the impact of rotary position embedding (RoPE) on proposed outlier-aware rotation and further smooths outliers across heads; (iii) Attention-Sink-Aware Quantization, which leverages the massive activations to precisely identify and protect attention sinks. RotateKV achieves less than 0.3 perplexity (PPL) degradation with 2-bit quantization on WikiText-2 using LLaMA-2-13B, maintains strong CoT reasoning and long-context capabilities, with less than 1.7\% degradation on GSM8K, outperforming existing methods even at lower average bit-widths. RotateKV also showcases a 3.97x reduction in peak memory usage, supports 5.75x larger batch sizes, and achieves a 2.32x speedup in decoding stage.
Designing proteins with specific attributes offers an important solution to address biomedical challenges. Pre-trained protein large language models (LLMs) have shown promising results on protein sequence generation. However, to control sequence generation for specific attributes, existing work still exhibits poor functionality and structural stability. In this paper, we propose a novel controllable protein design method called CtrlProt. We finetune a protein LLM with a new multi-listwise preference optimization strategy to improve generation quality and support multi-attribute controllable generation. Experiments demonstrate that CtrlProt can meet functionality and structural stability requirements effectively, achieving state-of-the-art performance in both single-attribute and multi-attribute protein sequence generation.
Detecting vulnerability fix commits in open-source software is crucial for maintaining software security. To help OSS identify vulnerability fix commits, several automated approaches are developed. However, existing approaches like VulFixMiner and CoLeFunDa, focus solely on code changes, neglecting essential context from development artifacts. Tools like Vulcurator, which integrates issue reports, fail to leverage semantic associations between different development artifacts (e.g., pull requests and history vulnerability fixes). Moreover, they miss vulnerability fixes in tangled commits and lack explanations, limiting practical use. Hence to address those limitations, we propose LLM4VFD, a novel framework that leverages Large Language Models (LLMs) enhanced with Chain-of-Thought reasoning and In-Context Learning to improve the accuracy of vulnerability fix detection. LLM4VFD comprises three components: (1) Code Change Intention, which analyzes commit summaries, purposes, and implications using Chain-of-Thought reasoning; (2) Development Artifact, which incorporates context from related issue reports and pull requests; (3) Historical Vulnerability, which retrieves similar past vulnerability fixes to enrich context. More importantly, on top of the prediction, LLM4VFD also provides a detailed analysis and explanation to help security experts understand the rationale behind the decision. We evaluated LLM4VFD against state-of-the-art techniques, including Pre-trained Language Model-based approaches and vanilla LLMs, using a newly collected dataset, BigVulFixes. Experimental results demonstrate that LLM4VFD significantly outperforms the best-performed existing approach by 68.1%--145.4%. Furthermore, We conducted a user study with security experts, showing that the analysis generated by LLM4VFD improves the efficiency of vulnerability fix identification.
Can LLMs pick up language structure from examples? Evidence in prior work seems to indicate yes, as pretrained models repeatedly demonstrate the ability to adapt to new language structures and vocabularies. However, this line of research typically considers languages that are present within common pretraining datasets, or otherwise share notable similarities with these seen languages. In contrast, in this work we attempt to measure models' language understanding capacity while circumventing the risk of dataset recall. We parameterize large families of language tasks recognized by deterministic finite automata (DFAs), and can thus sample novel language reasoning problems to fairly evaulate LLMs regardless of training data. We find that, even in the strikingly simple setting of 3-state DFAs, LLMs underperform unparameterized ngram models on both language recognition and synthesis tasks. These results suggest that LLMs struggle to match the ability of basic language models in recognizing and reasoning over languages that are sufficiently distinct from the ones they see at training time, underscoring the distinction between learning individual languages and possessing a general theory of language.
One of the most widely used methods to evaluate LLMs are Multiple Choice Question (MCQ) tests. MCQ benchmarks enable the testing of LLM knowledge on almost any topic at scale as the results can be processed automatically. To help the LLM answer, a few examples called few shots can be included in the prompt. Moreover, the LLM can be asked to answer the question directly with the selected option or to first provide the reasoning and then the selected answer, which is known as chain of thought. In addition to checking whether the selected answer is correct, the evaluation can look at the LLM-estimated probability of its response as an indication of the confidence of the LLM in the response. In this paper, we study how the LLM confidence in its answer depends on whether the model has been asked to answer directly or to provide the reasoning before answering. The results of the evaluation of questions on a wide range of topics in seven different models show that LLMs are more confident in their answers when they provide reasoning before the answer. This occurs regardless of whether the selected answer is correct. Our hypothesis is that this behavior is due to the reasoning that modifies the probability of the selected answer, as the LLM predicts the answer based on the input question and the reasoning that supports the selection made. Therefore, LLM estimated probabilities seem to have intrinsic limitations that should be understood in order to use them in evaluation procedures. Interestingly, the same behavior has been observed in humans, for whom explaining an answer increases confidence in its correctness.
Modern datasets often consist of numerous samples with abundant features and associated timestamps. Analyzing such datasets to uncover underlying events typically requires complex statistical methods and substantial domain expertise. A notable example, and the primary data focus of this paper, is the global synthetic dataset from the Counter Trafficking Data Collaborative (CTDC) -- a global hub of human trafficking data containing over 200,000 anonymized records spanning from 2002 to 2022, with numerous categorical features for each record. In this paper, we propose a fast and scalable method for analyzing and extracting significant categorical feature interactions, and querying large language models (LLMs) to generate data-driven insights that explain these interactions. Our approach begins with a binarization step for categorical features using one-hot encoding, followed by the computation of graph covariance at each time. This graph covariance quantifies temporal changes in dependence structures within categorical data and is established as a consistent dependence measure under the Bernoulli distribution. We use this measure to identify significant feature pairs, such as those with the most frequent trends over time or those exhibiting sudden spikes in dependence at specific moments. These extracted feature pairs, along with their timestamps, are subsequently passed to an LLM tasked with generating potential explanations of the underlying events driving these dependence changes. The effectiveness of our method is demonstrated through extensive simulations, and its application to the CTDC dataset reveals meaningful feature pairs and potential data stories underlying the observed feature interactions.
In knowledge-intensive tasks, especially in high-stakes domains like medicine and law, it is critical not only to retrieve relevant information but also to provide causal reasoning and explainability. Large language models (LLMs) have achieved remarkable performance in natural language understanding and generation tasks. However, they often suffer from limitations such as difficulty in incorporating new knowledge, generating hallucinations, and explaining their reasoning process. To address these challenges, integrating knowledge graphs with Graph Retrieval-Augmented Generation (Graph RAG) has emerged as an effective solution. Traditional Graph RAG methods often rely on simple graph traversal or semantic similarity, which do not capture causal relationships or align well with the model's internal reasoning steps. This paper proposes a novel pipeline that filters large knowledge graphs to emphasize cause-effect edges, aligns the retrieval process with the model's chain-of-thought (CoT), and enhances reasoning through multi-stage path improvements. Experiments on medical question-answering tasks show consistent gains, with up to a 10\% absolute improvement across multiple large language models (LLMs). This approach demonstrates the value of combining causal reasoning with stepwise retrieval, leading to more interpretable and logically grounded solutions for complex queries.
Large language models (LLMs) have excelled in various applications, yet serving them at scale is challenging due to their substantial resource demands and high latency. Our real-world studies reveal that over 60% of user requests to LLMs have semantically similar counterparts, suggesting the potential for knowledge sharing among requests. However, naively caching and reusing past responses leads to large quality degradation. In this paper, we introduce EchoLM, an in-context caching system that leverages historical requests as examples to guide response generation, enabling selective offloading of requests to more efficient LLMs. However, enabling this real-time knowledge transfer leads to intricate tradeoffs between response quality, latency, and system throughput at scale. For a new request, EchoLM identifies similar, high-utility examples and efficiently prepends them to the input for better response. At scale, EchoLM adaptively routes requests to LLMs of varying capabilities, accounting for response quality and serving loads. EchoLM employs a cost-aware cache replay mechanism to improve example quality and coverage offline, maximizing cache utility and runtime efficiency. Evaluations on millions of open-source requests demonstrate that EchoLM has a throughput improvement of 1.4-5.9x while reducing latency by 28-71% without hurting response quality on average.
Equitable access to reliable health information is vital for public health, but the quality of online health resources varies by language, raising concerns about inconsistencies in Large Language Models (LLMs) for healthcare. In this study, we examine the consistency of responses provided by LLMs to health-related questions across English, German, Turkish, and Chinese. We largely expand the HealthFC dataset by categorizing health-related questions by disease type and broadening its multilingual scope with Turkish and Chinese translations. We reveal significant inconsistencies in responses that could spread healthcare misinformation. Our main contributions are 1) a multilingual health-related inquiry dataset with meta-information on disease categories, and 2) a novel prompt-based evaluation workflow that enables sub-dimensional comparisons between two languages through parsing. Our findings highlight key challenges in deploying LLM-based tools in multilingual contexts and emphasize the need for improved cross-lingual alignment to ensure accurate and equitable healthcare information.
As Industry 4.0 transforms the food industry, the role of software in achieving compliance with food-safety regulations is becoming increasingly critical. Food-safety regulations, like those in many legal domains, have largely been articulated in a technology-independent manner to ensure their longevity and broad applicability. However, this approach leaves a gap between the regulations and the modern systems and software increasingly used to implement them. In this article, we pursue two main goals. First, we conduct a Grounded Theory study of food-safety regulations and develop a conceptual characterization of food-safety concepts that closely relate to systems and software requirements. Second, we examine the effectiveness of two families of large language models (LLMs) -- BERT and GPT -- in automatically classifying legal provisions based on requirements-related food-safety concepts. Our results show that: (a) when fine-tuned, the accuracy differences between the best-performing models in the BERT and GPT families are relatively small. Nevertheless, the most powerful model in our experiments, GPT-4o, still achieves the highest accuracy, with an average Precision of 89% and an average Recall of 87%; (b) few-shot learning with GPT-4o increases Recall to 97% but decreases Precision to 65%, suggesting a trade-off between fine-tuning and few-shot learning; (c) despite our training examples being drawn exclusively from Canadian regulations, LLM-based classification performs consistently well on test provisions from the US, indicating a degree of generalizability across regulatory jurisdictions; and (d) for our classification task, LLMs significantly outperform simpler baselines constructed using long short-term memory (LSTM) networks and automatic keyword extraction.
Recent large language models (LLMs) have demonstrated significant advancements, particularly in their ability to serve as agents thereby surpassing their traditional role as chatbots. These agents can leverage their planning and tool utilization capabilities to address tasks specified at a high level. However, a standardized dataset to benchmark the agent capabilities of LLMs in medical applications is currently lacking, making the evaluation of LLMs on complex tasks in interactive healthcare environments challenging. To address this gap, we introduce MedAgentBench, a broad evaluation suite designed to assess the agent capabilities of large language models within medical records contexts. MedAgentBench encompasses 100 patient-specific clinically-derived tasks from 10 categories written by human physicians, realistic profiles of 100 patients with over 700,000 data elements, a FHIR-compliant interactive environment, and an accompanying codebase. The environment uses the standard APIs and communication infrastructure used in modern EMR systems, so it can be easily migrated into live EMR systems. MedAgentBench presents an unsaturated agent-oriented benchmark that current state-of-the-art LLMs exhibit some ability to succeed at. The best model (GPT-4o) achieves a success rate of 72%. However, there is still substantial space for improvement to give the community a next direction to optimize. Furthermore, there is significant variation in performance across task categories. MedAgentBench establishes this and is publicly available at https://github.com/stanfordmlgroup/MedAgentBench , offering a valuable framework for model developers to track progress and drive continuous improvements in the agent capabilities of large language models within the medical domain.
Evaluating Large Language Models (LLMs) often requires costly human annotations. To address this, LLM-based judges have been proposed, which compare the outputs of two LLMs enabling the ranking of models without human intervention. While several approaches have been proposed, many confounding factors are present between different papers. For instance the model, the prompt and other hyperparameters are typically changed at the same time making apple-to-apple comparisons challenging. In this paper, we propose to systematically analyze and tune hyperparameter of LLM judges. To alleviate the high cost of evaluating a judge, we propose to leverage multi-objective multi-fidelity which allows to find judges that trades accuracy for cost and also reduce significantly the cost of the search. Our method identifies judges that not only outperform existing benchmarks in accuracy and cost-efficiency but also utilize open-weight models, ensuring greater accessibility and reproducibility.
Local search preprocessing makes Conflict-Driven Clause Learning (CDCL) solvers faster by providing high-quality starting points and modern SAT solvers have incorporated this technique into their preprocessing steps. However, these tools rely on basic strategies that miss the structural patterns in problems. We present a method that applies Large Language Models (LLMs) to analyze Python-based encoding code. This reveals hidden structural patterns in how problems convert into SAT. Our method automatically generates specialized local search algorithms that find these patterns and use them to create strong initial assignments. This works for any problem instance from the same encoding type. Our tests show encouraging results, achieving faster solving times compared to baseline preprocessing systems.
Citizen reporting platforms like Safe City in India help the public and authorities stay informed about sexual harassment incidents. However, the high volume of data shared on these platforms makes reviewing each individual case challenging. Therefore, a summarization algorithm capable of processing and understanding various Indian code-mixed languages is essential. In recent years, Large Language Models (LLMs) have shown exceptional performance in NLP tasks, including summarization. LLMs inherently produce abstractive summaries by paraphrasing the original text, while the generation of extractive summaries - selecting specific subsets from the original text - through LLMs remains largely unexplored. Moreover, LLMs have a limited context window size, restricting the amount of data that can be processed at once. We tackle these challenge by introducing LaMSUM, a novel multi-level framework designed to generate extractive summaries for large collections of Safe City posts using LLMs. LaMSUM integrates summarization with different voting methods to achieve robust summaries. Extensive evaluation using three popular LLMs (Llama, Mistral and GPT-4o) demonstrates that LaMSUM outperforms state-of-the-art extractive summarization methods for Safe City posts. Overall, this work represents one of the first attempts to achieve extractive summarization through LLMs, and is likely to support stakeholders by offering a comprehensive overview and enabling them to develop effective policies to minimize incidents of unwarranted harassment.
Bidders in combinatorial auctions face significant challenges when describing their preferences to an auctioneer. Classical work on preference elicitation focuses on query-based techniques inspired from proper learning--often via proxies that interface between bidders and an auction mechanism--to incrementally learn bidder preferences as needed to compute efficient allocations. Although such elicitation mechanisms enjoy theoretical query efficiency, the amount of communication required may still be too cognitively taxing in practice. We propose a family of efficient LLM-based proxy designs for eliciting preferences from bidders using natural language. Our proposed mechanism combines LLM pipelines and DNF-proper-learning techniques to quickly approximate preferences when communication is limited. To validate our approach, we create a testing sandbox for elicitation mechanisms that communicate in natural language. In our experiments, our most promising LLM proxy design reaches approximately efficient outcomes with five times fewer queries than classical proper learning based elicitation mechanisms.
Machine learning (ML) has the potential to revolutionize various domains, but its adoption is often hindered by the disconnect between the needs of domain experts and translating these needs into robust and valid ML tools. Despite recent advances in LLM-based co-pilots to democratize ML for non-technical domain experts, these systems remain predominantly focused on model-centric aspects while overlooking critical data-centric challenges. This limitation is problematic in complex real-world settings where raw data often contains complex issues, such as missing values, label noise, and domain-specific nuances requiring tailored handling. To address this we introduce CliMB-DC, a human-guided, data-centric framework for LLM co-pilots that combines advanced data-centric tools with LLM-driven reasoning to enable robust, context-aware data processing. At its core, CliMB-DC introduces a novel, multi-agent reasoning system that combines a strategic coordinator for dynamic planning and adaptation with a specialized worker agent for precise execution. Domain expertise is then systematically incorporated to guide the reasoning process using a human-in-the-loop approach. To guide development, we formalize a taxonomy of key data-centric challenges that co-pilots must address. Thereafter, to address the dimensions of the taxonomy, we integrate state-of-the-art data-centric tools into an extensible, open-source architecture, facilitating the addition of new tools from the research community. Empirically, using real-world healthcare datasets we demonstrate CliMB-DC's ability to transform uncurated datasets into ML-ready formats, significantly outperforming existing co-pilot baselines for handling data-centric challenges. CliMB-DC promises to empower domain experts from diverse domains -- healthcare, finance, social sciences and more -- to actively participate in driving real-world impact using ML.
We provide a systematic understanding of the impact of specific components and wordings used in prompts on the effectiveness of rankers based on zero-shot Large Language Models (LLMs). Several zero-shot ranking methods based on LLMs have recently been proposed. Among many aspects, methods differ across (1) the ranking algorithm they implement, e.g., pointwise vs. listwise, (2) the backbone LLMs used, e.g., GPT3.5 vs. FLAN-T5, (3) the components and wording used in prompts, e.g., the use or not of role-definition (role-playing) and the actual words used to express this. It is currently unclear whether performance differences are due to the underlying ranking algorithm, or because of spurious factors such as better choice of words used in prompts. This confusion risks to undermine future research. Through our large-scale experimentation and analysis, we find that ranking algorithms do contribute to differences between methods for zero-shot LLM ranking. However, so do the LLM backbones -- but even more importantly, the choice of prompt components and wordings affect the ranking. In fact, in our experiments, we find that, at times, these latter elements have more impact on the ranker's effectiveness than the actual ranking algorithms, and that differences among ranking methods become more blurred when prompt variations are considered.
We explore the potential of self-play training for large language models (LLMs) in a two-player adversarial language game called Adversarial Taboo. In this game, an attacker and a defender communicate around a target word only visible to the attacker. The attacker aims to induce the defender to speak the target word unconsciously, while the defender tries to infer the target word from the attacker's utterances. To win the game, both players must have sufficient knowledge about the target word and high-level reasoning ability to infer and express in this information-reserved conversation. Hence, we are curious about whether LLMs' reasoning ability can be further enhanced by Self-Playing this Adversarial language Game (SPAG). With this goal, we select several open-source LLMs and let each act as the attacker and play with a copy of itself as the defender on an extensive range of target words. Through reinforcement learning on the game outcomes, we observe that the LLMs' performances uniformly improve on a broad range of reasoning benchmarks. Furthermore, iteratively adopting this self-play process can continuously promote LLMs' reasoning abilities. The code is available at https://github.com/Linear95/SPAG.
A recent approach to neurosymbolic reasoning is to explicitly combine the strengths of large language models (LLMs) and symbolic solvers to tackle complex reasoning tasks. However, current approaches face significant limitations, including poor generalizability due to task-specific prompts, inefficiencies caused by the lack of separation between knowledge and queries, and restricted inferential capabilities. These shortcomings hinder their scalability and applicability across diverse domains. In this paper, we introduce VERUS-LM, a novel framework designed to address these challenges. VERUS-LM employs a generic prompting mechanism, clearly separates domain knowledge from queries, and supports a wide range of different logical reasoning tasks. This framework enhances adaptability, reduces computational cost, and allows for richer forms of reasoning, such as optimization and constraint satisfaction. We show that our approach succeeds in diverse reasoning on a novel dataset, markedly outperforming LLMs. Additionally, our system achieves competitive results on common reasoning benchmarks when compared to other state-of-the-art approaches, and significantly surpasses them on the difficult AR-LSAT dataset. By pushing the boundaries of hybrid reasoning, VERUS-LM represents a significant step towards more versatile neurosymbolic AI systems
Efficient control in long-horizon robotic manipulation is challenging due to complex representation and policy learning requirements. Model-based visual reinforcement learning (RL) has shown great potential in addressing these challenges but still faces notable limitations, particularly in handling sparse rewards and complex visual features in long-horizon environments. To address these limitations, we propose the Recognize-Sense-Plan-Act (RSPA) pipeline for long-horizon tasks and further introduce RoboHorizon, an LLM-assisted multi-view world model tailored for long-horizon robotic manipulation. In RoboHorizon, pre-trained LLMs generate dense reward structures for multi-stage sub-tasks based on task language instructions, enabling robots to better recognize long-horizon tasks. Keyframe discovery is then integrated into the multi-view masked autoencoder (MAE) architecture to enhance the robot's ability to sense critical task sequences, strengthening its multi-stage perception of long-horizon processes. Leveraging these dense rewards and multi-view representations, a robotic world model is constructed to efficiently plan long-horizon tasks, enabling the robot to reliably act through RL algorithms. Experiments on two representative benchmarks, RLBench and FurnitureBench, show that RoboHorizon outperforms state-of-the-art visual model-based RL methods, achieving a 23.35% improvement in task success rates on RLBench's 4 short-horizon tasks and a 29.23% improvement on 6 long-horizon tasks from RLBench and 3 furniture assembly tasks from FurnitureBench.
As software systems grow more complex, automated testing has become essential to ensuring reliability and performance. Traditional methods for boundary value test input generation can be time-consuming and may struggle to address all potential error cases effectively, especially in systems with intricate or highly variable boundaries. This paper presents a framework for assessing the effectiveness of large language models (LLMs) in generating boundary value test inputs for white-box software testing by examining their potential through prompt engineering. Specifically, we evaluate the effectiveness of LLM-based test input generation by analyzing fault detection rates and test coverage, comparing these LLM-generated test sets with those produced using traditional boundary value analysis methods. Our analysis shows the strengths and limitations of LLMs in boundary value generation, particularly in detecting common boundary-related issues. However, they still face challenges in certain areas, especially when handling complex or less common test inputs. This research provides insights into the role of LLMs in boundary value testing, underscoring both their potential and areas for improvement in automated testing methods.
Large language models (LLMs) often exhibit gender bias, posing challenges for their safe deployment. Existing methods to mitigate bias lack a comprehensive understanding of its mechanisms or compromise the model's core capabilities. To address these issues, we propose the CommonWords dataset, to systematically evaluate gender bias in LLMs. Our analysis reveals pervasive bias across models and identifies specific neuron circuits, including gender neurons and general neurons, responsible for this behavior. Notably, editing even a small number of general neurons can disrupt the model's overall capabilities due to hierarchical neuron interactions. Based on these insights, we propose an interpretable neuron editing method that combines logit-based and causal-based strategies to selectively target biased neurons. Experiments on five LLMs demonstrate that our method effectively reduces gender bias while preserving the model's original capabilities, outperforming existing fine-tuning and editing approaches. Our findings contribute a novel dataset, a detailed analysis of bias mechanisms, and a practical solution for mitigating gender bias in LLMs.
Large Language Models (LLMs) are widely applied to downstream domains. However, current LLMs for high-stakes domain tasks, such as financial investment and legal QA, typically generate brief answers without reasoning processes and explanations. This limits users' confidence in making decisions based on their responses. While original CoT shows promise, it lacks self-correction mechanisms during reasoning. This work introduces Domain$o1$s, which enhances LLMs' reasoning capabilities on domain tasks through supervised fine-tuning and tree search. We construct CoT-stock-2k and CoT-legal-2k datasets for fine-tuning models that activate domain-specific reasoning steps based on their judgment. Additionally, we propose Selective Tree Exploration to spontaneously explore solution spaces and sample optimal reasoning paths to improve performance. We also introduce PROOF-Score, a new metric for evaluating domain models' explainability, complementing traditional accuracy metrics with richer assessment dimensions. Extensive experiments on stock investment recommendation and legal reasoning QA tasks demonstrate Domaino1s's leading performance and explainability. Our code is available at https://anonymous.4open.science/r/Domaino1s-006F/.
The success of Large Language Models (LLMs) in various domains has led researchers to apply them to graph-related problems by converting graph data into natural language text. However, unlike graph data, natural language inherently has sequential order. We observe that when the order of nodes or edges in the natural language description of a graph is shuffled, despite describing the same graph, model performance fluctuates between high performance and random guessing. Additionally, due to the limited input context length of LLMs, current methods typically randomly sample neighbors of target nodes as representatives of their neighborhood, which may not always be effective for accurate reasoning. To address these gaps, we introduce GraphBC. This novel model framework features an Order Selector Module to ensure proper serialization order of the graph and a Subgraph Sampling Module to sample subgraphs with better structure for better reasoning. Furthermore, we propose Graph CoT obtained through distillation, and enhance LLM's reasoning and zero-shot learning capabilities for graph tasks through instruction tuning. Experiments on multiple datasets for node classification and graph question-answering demonstrate that GraphBC improves LLMs' performance and generalization ability on graph tasks.
Recent advances in eXplainable AI (XAI) for education have highlighted a critical challenge: ensuring that explanations for state-of-the-art AI models are understandable for non-technical users such as educators and students. In response, we introduce iLLuMinaTE, a zero-shot, chain-of-prompts LLM-XAI pipeline inspired by Miller's cognitive model of explanation. iLLuMinaTE is designed to deliver theory-driven, actionable feedback to students in online courses. iLLuMinaTE navigates three main stages - causal connection, explanation selection, and explanation presentation - with variations drawing from eight social science theories (e.g. Abnormal Conditions, Pearl's Model of Explanation, Necessity and Robustness Selection, Contrastive Explanation). We extensively evaluate 21,915 natural language explanations of iLLuMinaTE extracted from three LLMs (GPT-4o, Gemma2-9B, Llama3-70B), with three different underlying XAI methods (LIME, Counterfactuals, MC-LIME), across students from three diverse online courses. Our evaluation involves analyses of explanation alignment to the social science theory, understandability of the explanation, and a real-world user preference study with 114 university students containing a novel actionability simulation. We find that students prefer iLLuMinaTE explanations over traditional explainers 89.52% of the time. Our work provides a robust, ready-to-use framework for effectively communicating hybrid XAI-driven insights in education, with significant generalization potential for other human-centric fields.
We introduce DRESS, a novel approach for generating stylized large language model (LLM) responses through representation editing. Existing methods like prompting and fine-tuning are either insufficient for complex style adaptation or computationally expensive, particularly in tasks like NPC creation or character role-playing. Our approach leverages the over-parameterized nature of LLMs to disentangle a style-relevant subspace within the model's representation space to conduct representation editing, ensuring a minimal impact on the original semantics. By applying adaptive editing strengths, we dynamically adjust the steering vectors in the style subspace to maintain both stylistic fidelity and semantic integrity. We develop two stylized QA benchmark datasets to validate the effectiveness of DRESS, and the results demonstrate significant improvements compared to baseline methods such as prompting and ITI. In short, DRESS is a lightweight, train-free solution for enhancing LLMs with flexible and effective style control, making it particularly useful for developing stylized conversational agents. Codes and benchmark datasets are available at https://github.com/ArthurLeoM/DRESS-LLM.
We present FireRedASR, a family of large-scale automatic speech recognition (ASR) models for Mandarin, designed to meet diverse requirements in superior performance and optimal efficiency across various applications. FireRedASR comprises two variants: FireRedASR-LLM: Designed to achieve state-of-the-art (SOTA) performance and to enable seamless end-to-end speech interaction. It adopts an Encoder-Adapter-LLM framework leveraging large language model (LLM) capabilities. On public Mandarin benchmarks, FireRedASR-LLM (8.3B parameters) achieves an average Character Error Rate (CER) of 3.05%, surpassing the latest SOTA of 3.33% with an 8.4% relative CER reduction (CERR). It demonstrates superior generalization capability over industrial-grade baselines, achieving 24%-40% CERR in multi-source Mandarin ASR scenarios such as video, live, and intelligent assistant. FireRedASR-AED: Designed to balance high performance and computational efficiency and to serve as an effective speech representation module in LLM-based speech models. It utilizes an Attention-based Encoder-Decoder (AED) architecture. On public Mandarin benchmarks, FireRedASR-AED (1.1B parameters) achieves an average CER of 3.18%, slightly worse than FireRedASR-LLM but still outperforming the latest SOTA model with over 12B parameters. It offers a more compact size, making it suitable for resource-constrained applications. Moreover, both models exhibit competitive results on Chinese dialects and English speech benchmarks and excel in singing lyrics recognition. To advance research in speech processing, we release our models and inference code at https://github.com/FireRedTeam/FireRedASR.
Decision-making agents based on pre-trained Large Language Models (LLMs) are increasingly being deployed across various domains of human activity. While their applications are currently rather specialized, several research efforts are under way to develop more generalist agents. As LLM-based systems become more agentic, their influence on human activity will grow and the transparency of this will decrease. Consequently, developing effective methods for aligning them to human values is vital. The prevailing practice in alignment often relies on human preference data (e.g., in RLHF or DPO), in which values are implicit and are essentially deduced from relative preferences over different model outputs. In this work, instead of relying on human feedback, we introduce the design of reward functions that explicitly encode core human values for Reinforcement Learning-based fine-tuning of foundation agent models. Specifically, we use intrinsic rewards for the moral alignment of LLM agents. We evaluate our approach using the traditional philosophical frameworks of Deontological Ethics and Utilitarianism, quantifying moral rewards for agents in terms of actions and consequences on the Iterated Prisoner's Dilemma (IPD) environment. We also show how moral fine-tuning can be deployed to enable an agent to unlearn a previously developed selfish strategy. Finally, we find that certain moral strategies learned on the IPD game generalize to several other matrix game environments. In summary, we demonstrate that fine-tuning with intrinsic rewards is a promising general solution for aligning LLM agents to human values, and it might represent a more transparent and cost-effective alternative to currently predominant alignment techniques.
Code summarization aims to generate concise natural language descriptions for source code. Deep learning has been used more and more recently in software engineering, particularly for tasks like code creation and summarization. Specifically, it appears that the most current Large Language Models with coding perform well on these tasks. Large Language Models (LLMs) have significantly advanced the field of code summarization, providing sophisticated methods for generating concise and accurate summaries of source code. This study aims to perform a comparative analysis of several open-source LLMs, namely LLaMA-3, Phi-3, Mistral, and Gemma. These models' performance is assessed using important metrics such as BLEU\textsubscript{3.1} and ROUGE\textsubscript{3.2}. Through this analysis, we seek to identify the strengths and weaknesses of each model, offering insights into their applicability and effectiveness in code summarization tasks. Our findings contribute to the ongoing development and refinement of LLMs, supporting their integration into tools that enhance software development and maintenance processes.
High annotation costs from hiring or crowdsourcing complicate the creation of large, high-quality datasets needed for training reliable text classifiers. Recent research suggests using Large Language Models (LLMs) to automate the annotation process, reducing these costs while maintaining data quality. LLMs have shown promising results in annotating downstream tasks like hate speech detection and political framing. Building on the success in these areas, this study investigates whether LLMs are viable for annotating the complex task of media bias detection and whether a downstream media bias classifier can be trained on such data. We create annolexical, the first large-scale dataset for media bias classification with over 48000 synthetically annotated examples. Our classifier, fine-tuned on this dataset, surpasses all of the annotator LLMs by 5-9 percent in Matthews Correlation Coefficient (MCC) and performs close to or outperforms the model trained on human-labeled data when evaluated on two media bias benchmark datasets (BABE and BASIL). This study demonstrates how our approach significantly reduces the cost of dataset creation in the media bias domain and, by extension, the development of classifiers, while our subsequent behavioral stress-testing reveals some of its current limitations and trade-offs.
Maintaining consistent model performance across domains is a fundamental challenge in machine learning. While recent work has explored using LLM-generated data for fine-tuning, its impact on cross-domain generalization remains poorly understood. In this paper, we present a systematic analysis revealing that fine-tuning with LLM-generated data not only improves target task performance but also reduces out-of-domain (OOD) degradation compared to fine-tuning with ground truth data. Through analyzing the data sequence in tasks of various domains, we demonstrate that this enhanced OOD robustness stems from a reduced prevalence of high perplexity tokens in LLM-generated sequences. Following this hypothesis we showed that masking high perplexity tokens in ground truth training data also achieves similar OOD preservation comparable to using LLM-generated data. Extensive experiments across diverse model architectures and scales, including Gemma2-2B, Mistral-7B and Llama3-8B, corroborate the consistency of our findings. To the best of our knowledge, this work provides the first mechanistic explanation for the superior OOD robustness conferred by LLM-generated training data, offering valuable insights for developing more robust fine-tuning strategies.
COAST: Enhancing the Code Debugging Ability of LLMs through Communicative Agent Based Data Synthesis
Code debugging is a vital stage of software development, essential for ensuring the reliability and performance of Large Language Models (LLMs) in code generation task. Human debugging typically follows a multi-stage process, which includes Bug Localization, Bug Identification, Code Repair, and Code Recognition. However, existing code debugging benchmarks predominantly focus on the Code Repair stage, which offers only a limited perspective on evaluating the debugging capabilities of LLMs. In this paper, we introduce DEBUGEVAL, a comprehensive benchmark for evaluating the debugging abilities of LLMs by emulating the multi-stage human debugging process. Through evaluating on DEBUGEVAL, we observe that 7B-scale models consistently underperform compared to their larger counterparts, highlighting their limitations in comprehending code semantics. In this case, we propose the COmmunicative Agent-based data SynThesis (COAST) framework, which employs a multi-agent system to generate high-quality training data for supervised fine-tuning (SFT). Experimental results demonstrate that COAST-generated data outperform human-curated and GPT-4-generated data, enabling 7B-scale LLMs to achieve debugging performance comparable to GPT-3.5.
Large language model (LLM) inference workload dominates a wide variety of modern AI applications, ranging from multi-turn conversation to document analysis. Balancing fairness and efficiency is critical for managing diverse client workloads with varying prefix patterns. Unfortunately, existing fair scheduling algorithms for LLM serving, such as Virtual Token Counter (VTC), fail to take prefix locality into consideration and thus suffer from poor performance. On the other hand, locality-aware scheduling algorithms in existing LLM serving frameworks tend to maximize the prefix cache hit rate without considering fair sharing among clients. This paper introduces the first locality-aware fair scheduling algorithm, Deficit Longest Prefix Match (DLPM), which can maintain a high degree of prefix locality with a fairness guarantee. We also introduce a novel algorithm, Double Deficit LPM (D$^2$LPM), extending DLPM for the distributed setup that can find a balance point among fairness, locality, and load-balancing. Our extensive evaluation demonstrates the superior performance of DLPM and D$^2$LPM in ensuring fairness while maintaining high throughput (up to 2.87$\times$ higher than VTC) and low per-client (up to 7.18$\times$ lower than state-of-the-art distributed LLM serving system) latency.
Automated grading has become an essential tool in education technology due to its ability to efficiently assess large volumes of student work, provide consistent and unbiased evaluations, and deliver immediate feedback to enhance learning. However, current systems face significant limitations, including the need for large datasets in few-shot learning methods, a lack of personalized and actionable feedback, and an overemphasis on benchmark performance rather than student experience. To address these challenges, we propose a Zero-Shot Large Language Model (LLM)-Based Automated Assignment Grading (AAG) system. This framework leverages prompt engineering to evaluate both computational and explanatory student responses without requiring additional training or fine-tuning. The AAG system delivers tailored feedback that highlights individual strengths and areas for improvement, thereby enhancing student learning outcomes. Our study demonstrates the system's effectiveness through comprehensive evaluations, including survey responses from higher education students that indicate significant improvements in motivation, understanding, and preparedness compared to traditional grading methods. The results validate the AAG system's potential to transform educational assessment by prioritizing learning experiences and providing scalable, high-quality feedback.
Large Language Models (LLM) are increasingly being explored for problem-solving tasks. However, their strategic planning capability is often viewed with skepticism. Recent studies have incorporated the Monte Carlo Tree Search (MCTS) algorithm to augment the planning capacity of LLM. Despite its potential, MCTS relies on extensive sampling simulations to approximate the true reward distribution, leading to two primary issues. Firstly, MCTS is effective for tasks like the Game of Go, where simulation results can yield objective rewards (e.g., 1 for a win and 0 for a loss). However, for tasks such as question answering, the result of a simulation is the answer to the question, which cannot obtain an objective reward without the ground truth. Secondly, obtaining statistically significant reward estimations typically requires a sample size exceeding 30 simulations, resulting in excessive token usage and time consumption. To address these challenges, we present Multi-Agent System with Tactical Execution and Reasoning using LLM Specialized MCTS (MASTER), a novel framework that coordinates agent recruitment and communication using LLM specialized MCTS. This system autonomously adjusts the number of agents based on task complexity and ensures focused communication among them. Comprehensive experiments across various tasks demonstrate the effectiveness of our proposed framework. It achieves 76% accuracy on HotpotQA and 80% on WebShop, setting new state-of-the-art performance on these datasets.
Large Language Models (LLMs) changed the way we design and interact with software systems. Their ability to process and extract information from text has drastically improved productivity in a number of routine tasks. Developers that want to include these models in their software stack, however, face a dreadful challenge: debugging LLMs' inconsistent behavior across minor variations of the prompt. We therefore introduce two metrics for classification tasks, namely sensitivity and consistency, which are complementary to task performance. First, sensitivity measures changes of predictions across rephrasings of the prompt, and does not require access to ground truth labels. Instead, consistency measures how predictions vary across rephrasings for elements of the same class. We perform an empirical comparison of these metrics on text classification tasks, using them as guideline for understanding failure modes of the LLM. Our hope is that sensitivity and consistency will be helpful to guide prompt engineering and obtain LLMs that balance robustness with performance.
Temporal point processes (TPPs) are stochastic process models used to characterize event sequences occurring in continuous time. Traditional statistical TPPs have a long-standing history, with numerous models proposed and successfully applied across diverse domains. In recent years, advances in deep learning have spurred the development of neural TPPs, enabling greater flexibility and expressiveness in capturing complex temporal dynamics. The emergence of large language models (LLMs) has further sparked excitement, offering new possibilities for modeling and analyzing event sequences by leveraging their rich contextual understanding. This survey presents a comprehensive review of recent research on TPPs from three perspectives: Bayesian, deep learning, and LLM approaches. We begin with a review of the fundamental concepts of TPPs, followed by an in-depth discussion of model design and parameter estimation techniques in these three frameworks. We also revisit classic application areas of TPPs to highlight their practical relevance. Finally, we outline challenges and promising directions for future research.
Advances in Large Language Models (LLMs) have sparked interest in their ability to solve Olympiad-level math problems. However, the training and evaluation of these models are constrained by the limited size and quality of available datasets, as creating large-scale data for such advanced problems requires extensive effort from human experts. In addition, current benchmarks are prone to contamination, leading to unreliable evaluations. In this paper, we present an automated pipeline that leverages the rich resources of the Art of Problem Solving (AoPS) forum, which predominantly features Olympiad-level problems and community-driven solutions. Using open-source LLMs, we develop a method to extract question-answer pairs from the forum, resulting in AoPS-Instruct, a dataset of more than 600,000 high-quality QA pairs. Our experiments demonstrate that fine-tuning LLMs on AoPS-Instruct improves their reasoning abilities across various benchmarks. Moreover, we build an automatic pipeline that introduces LiveAoPSBench, an evolving evaluation set with timestamps, derived from the latest forum data, providing a contamination-resistant benchmark for assessing LLM performance. Notably, we observe a significant decline in LLM performance over time, suggesting their success on older examples may stem from pre-training exposure rather than true reasoning ability. Our work presents a scalable approach to creating and maintaining large-scale, high-quality datasets for advanced math reasoning, offering valuable insights into the capabilities and limitations of LLMs in this domain. Our benchmark and code is available at https://github.com/DSL-Lab/aops
Graph databases like Neo4j are gaining popularity for handling complex, interconnected data, over traditional relational databases in modeling and querying relationships. While translating natural language into SQL queries is well-researched, generating Cypher queries for Neo4j remains relatively underexplored. In this work, we present an automated, LLM-Supervised, pipeline to generate high-quality synthetic data for Text2Cypher. Our Cypher data generation pipeline introduces LLM-As-Database-Filler, a novel strategy for ensuring Cypher query correctness, thus resulting in high quality generations. Using our pipeline, we generate high quality Text2Cypher data - SynthCypher containing 29.8k instances across various domains and queries with varying complexities. Training open-source LLMs like LLaMa-3.1-8B, Mistral-7B, and QWEN-7B on SynthCypher results in performance gains of up to 40% on the Text2Cypher test split and 30% on the SPIDER benchmark, adapted for graph databases.
Leveraging sparsity is crucial for optimizing large language model inference. however, modern LLMs employing SiLU as their activation function exhibit minimal activation sparsity. Recent research has proposed replacing SiLU with ReLU to induce significant activation sparsity and showed no downstream task accuracy degradation through fine tuning. However, taking full advantage of it required training a predictor to estimate this sparsity. In this paper, we introduce SparseInfer, a simple, light weight, and training free predictor for activation sparsity of ReLU field LLMs, in which activation sparsity is predicted by comparing only the sign bits of inputs and weights. To compensate for possible prediction inaccuracy, an adaptive tuning of the predictor's conservativeness is enabled, which can also serve as a control knob for optimizing LLM inference. The proposed method achieves approximately faster inference speed over the state of the art, with negligible accuracy loss of within 1%p.