llm - 2024_03
Navigation
Papers
FlashAttention (Dao, 2023) effectively reduces the quadratic peak memory usage to linear in training transformer-based large language models (LLMs) on a single GPU. In this paper, we introduce DISTFLASHATTN, a distributed memory-efficient attention mechanism optimized for long-context LLMs training. We propose three key techniques: token-level workload balancing, overlapping key-value communication, and a rematerialization-aware gradient checkpointing algorithm. We evaluate DISTFLASHATTN on Llama-7B and variants with sequence lengths from 32K to 512K. DISTFLASHATTN achieves 8x longer sequences, 4.45 - 5.64x speedup compared to Ring Self-Attention, 2 - 8x longer sequences, 1.24 - 2.01x speedup compared to Megatron-LM with FlashAttention. It achieves 1.67x and 1.26 - 1.88x speedup compared to recent Ring Attention and DeepSpeed-Ulysses. Code is available at https://github.com/RulinShao/LightSeq.
Single document news summarization has seen substantial progress on faithfulness in recent years, driven by research on the evaluation of factual consistency, or hallucinations. We ask whether these advances carry over to other text summarization domains. We propose a new evaluation benchmark on topic-focused dialogue summarization, generated by LLMs of varying sizes. We provide binary sentence-level human annotations of the factual consistency of these summaries along with detailed explanations of factually inconsistent sentences. Our analysis shows that existing LLMs hallucinate significant amounts of factual errors in the dialogue domain, regardless of the model's size. On the other hand, when LLMs, including GPT-4, serve as binary factual evaluators, they perform poorly and can be outperformed by prevailing state-of-the-art specialized factuality evaluation metrics. Finally, we conducted an analysis of hallucination types with a curated error taxonomy. We find that there are diverse errors and error distributions in model-generated summaries and that non-LLM based metrics can capture all error types better than LLM-based evaluators.
Recent advancements in open vocabulary models, like CLIP, have notably advanced zero-shot classification and segmentation by utilizing natural language for class-specific embeddings. However, most research has focused on improving model accuracy through prompt engineering, prompt learning, or fine-tuning with limited labeled data, thereby overlooking the importance of refining the class descriptors. This paper introduces a new approach to text-supervised semantic segmentation using supervision by a large language model (LLM) that does not require extra training. Our method starts from an LLM, like GPT-3, to generate a detailed set of subclasses for more accurate class representation. We then employ an advanced text-supervised semantic segmentation model to apply the generated subclasses as target labels, resulting in diverse segmentation results tailored to each subclass's unique characteristics. Additionally, we propose an assembly that merges the segmentation maps from the various subclass descriptors to ensure a more comprehensive representation of the different aspects in the test images. Through comprehensive experiments on three standard benchmarks, our method outperforms traditional text-supervised semantic segmentation methods by a marked margin.
In this study, we propose a novel human-like memory architecture designed for enhancing the cognitive abilities of large language model based dialogue agents. Our proposed architecture enables agents to autonomously recall memories necessary for response generation, effectively addressing a limitation in the temporal cognition of LLMs. We adopt the human memory cue recall as a trigger for accurate and efficient memory recall. Moreover, we developed a mathematical model that dynamically quantifies memory consolidation, considering factors such as contextual relevance, elapsed time, and recall frequency. The agent stores memories retrieved from the user's interaction history in a database that encapsulates each memory's content and temporal context. Thus, this strategic storage allows agents to recall specific memories and understand their significance to the user in a temporal context, similar to how humans recognize and recall past experiences.
Language models pre-trained on general text have achieved impressive results in diverse fields. Yet, the distinct linguistic characteristics of task-oriented dialogues (TOD) compared to general text limit the practical utility of existing language models. Current task-oriented dialogue pre-training methods overlook the one-to-many property of conversations, where multiple responses can be appropriate given the same conversation context. In this paper, we propose a novel dialogue pre-training model called DivTOD, which collaborates with LLMs to learn diverse task-oriented dialogue representations. DivTOD guides LLMs in transferring diverse knowledge to smaller models while removing domain knowledge that contradicts task-oriented dialogues. Experiments show that our model outperforms strong TOD baselines on various downstream dialogue tasks and learns the intrinsic diversity of task-oriented dialogues.
In this paper, we introduce a black-box prompt optimization method that uses an attacker LLM agent to uncover higher levels of memorization in a victim agent, compared to what is revealed by prompting the target model with the training data directly, which is the dominant approach of quantifying memorization in LLMs. We use an iterative rejection-sampling optimization process to find instruction-based prompts with two main characteristics: (1) minimal overlap with the training data to avoid presenting the solution directly to the model, and (2) maximal overlap between the victim model's output and the training data, aiming to induce the victim to spit out training data. We observe that our instruction-based prompts generate outputs with 23.7% higher overlap with training data compared to the baseline prefix-suffix measurements. Our findings show that (1) instruction-tuned models can expose pre-training data as much as their base-models, if not more so, (2) contexts other than the original training data can lead to leakage, and (3) using instructions proposed by other LLMs can open a new avenue of automated attacks that we should further study and explore. The code can be found at https://github.com/Alymostafa/Instruction_based_attack .
Skeleton-based action recognition has attracted lots of research attention. Recently, to build an accurate skeleton-based action recognizer, a variety of works have been proposed. Among them, some works use large model architectures as backbones of their recognizers to boost the skeleton data representation capability, while some other works pre-train their recognizers on external data to enrich the knowledge. In this work, we observe that large language models which have been extensively used in various natural language processing tasks generally hold both large model architectures and rich implicit knowledge. Motivated by this, we propose a novel LLM-AR framework, in which we investigate treating the Large Language Model as an Action Recognizer. In our framework, we propose a linguistic projection process to project each input action signal (i.e., each skeleton sequence) into its ``sentence format'' (i.e., an ``action sentence''). Moreover, we also incorporate our framework with several designs to further facilitate this linguistic projection process. Extensive experiments demonstrate the efficacy of our proposed framework.
MindScape aims to study the benefits of integrating time series behavioral patterns (e.g., conversational engagement, sleep, location) with Large Language Models (LLMs) to create a new form of contextual AI journaling, promoting self-reflection and well-being. We argue that integrating behavioral sensing in LLMs will likely lead to a new frontier in AI. In this Late-Breaking Work paper, we discuss the MindScape contextual journal App design that uses LLMs and behavioral sensing to generate contextual and personalized journaling prompts crafted to encourage self-reflection and emotional development. We also discuss the MindScape study of college students based on a preliminary user study and our upcoming study to assess the effectiveness of contextual AI journaling in promoting better well-being on college campuses. MindScape represents a new application class that embeds behavioral intelligence in AI.
With the rise of large language models (LLMs), ensuring they embody the principles of being helpful, honest, and harmless (3H), known as Human Alignment, becomes crucial. While existing alignment methods like RLHF, DPO, etc., effectively fine-tune LLMs to match preferences in the preference dataset, they often lead LLMs to highly receptive human input and external evidence, even when this information is poisoned. This leads to a tendency for LLMs to be Adaptive Chameleons when external evidence conflicts with their parametric memory. This exacerbates the risk of LLM being attacked by external poisoned data, which poses a significant security risk to LLM system applications such as Retrieval-augmented generation (RAG). To address the challenge, we propose a novel framework: Dialectical Alignment (DA), which (1) utilizes AI feedback to identify optimal strategies for LLMs to navigate inter-context conflicts and context-memory conflicts with different external evidence in context window (i.e., different ratios of poisoned factual contexts); (2) constructs the SFT dataset as well as the preference dataset based on the AI feedback and strategies above; (3) uses the above datasets for LLM alignment to defense poisoned context attack while preserving the effectiveness of in-context knowledge editing. Our experiments show that the dialectical alignment model improves poisoned data attack defense by 20 and does not require any additional prompt engineering or prior declaration of ``you may be attacked`` to the LLMs' context window.
Information extraction (IE) is a fundamental area in natural language processing where prompting large language models (LLMs), even with in-context examples, cannot defeat small LMs tuned on very small IE datasets. We observe that IE tasks, such as named entity recognition and relation extraction, all focus on extracting important information, which can be formalized as a label-to-span matching. In this paper, we propose a novel framework MetaIE to build a small LM as meta-model by learning to extract "important information", i.e., the meta-understanding of IE, so that this meta-model can be adapted to all kind of IE tasks effectively and efficiently. Specifically, MetaIE obtains the small LM via a symbolic distillation from an LLM following the label-to-span scheme. We construct the distillation dataset via sampling sentences from language model pre-training datasets (e.g., OpenWebText in our implementation) and prompting an LLM to identify the typed spans of "important information". We evaluate the meta-model under the few-shot adaptation setting. Extensive results on 13 datasets from 6 IE tasks confirm that MetaIE can offer a better starting point for few-shot tuning on IE datasets and outperform other meta-models from (1) vanilla language model pre-training, (2) multi-IE-task pre-training with human annotations, and (3) single-IE-task symbolic distillation from LLM. Moreover, we provide comprehensive analyses of MetaIE, such as the size of the distillation dataset, the meta-model architecture, and the size of the meta-model.
Large language models (LLMs) are highly adept at question answering and reasoning tasks, but when reasoning in a situational context, human expectations vary depending on the relevant cultural common ground. As languages are associated with diverse cultures, LLMs should also be culturally-diverse reasoners. In this paper, we study the ability of a wide range of state-of-the-art multilingual LLMs (mLLMs) to reason with proverbs and sayings in a conversational context. Our experiments reveal that: (1) mLLMs "know" limited proverbs and memorizing proverbs does not mean understanding them within a conversational context; (2) mLLMs struggle to reason with figurative proverbs and sayings, and when asked to select the wrong answer (instead of asking it to select the correct answer); and (3) there is a "culture gap" in mLLMs when reasoning about proverbs and sayings translated from other languages. We construct and release our evaluation dataset MAPS (MulticultrAl Proverbs and Sayings) for proverb understanding with conversational context for six different languages.
With ChatGPT's release, conversational prompting has become the most popular form of human-LLM interaction. However, its effectiveness is limited for more complex tasks involving reasoning, creativity, and iteration. Through a systematic analysis of HCI papers published since 2021, we identified four key phases in the human-LLM interaction flow - planning, facilitating, iterating, and testing - to precisely understand the dynamics of this process. Additionally, we have developed a taxonomy of four primary interaction modes: Mode 1: Standard Prompting, Mode 2: User Interface, Mode 3: Context-based, and Mode 4: Agent Facilitator. This taxonomy was further enriched using the "5W1H" guideline method, which involved a detailed examination of definitions, participant roles (Who), the phases that happened (When), human objectives and LLM abilities (What), and the mechanics of each interaction mode (How). We anticipate this taxonomy will contribute to the future design and evaluation of human-LLM interaction.
Large Language Models (LLMs) have demonstrated exceptional capabilities in various natural language tasks, often achieving performances that surpass those of humans. Despite these advancements, the domain of mathematics presents a distinctive challenge, primarily due to its specialized structure and the precision it demands. In this study, we adopted a two-step approach for investigating the proficiency of LLMs in answering mathematical questions. First, we employ the most effective LLMs, as identified by their performance on math question-answer benchmarks, to generate answers to 78 questions from the Math Stack Exchange (MSE). Second, a case analysis is conducted on the LLM that showed the highest performance, focusing on the quality and accuracy of its answers through manual evaluation. We found that GPT-4 performs best (nDCG of 0.48 and P@10 of 0.37) amongst existing LLMs fine-tuned for answering mathematics questions and outperforms the current best approach on ArqMATH3 Task1, considering P@10. Our Case analysis indicates that while the GPT-4 can generate relevant responses in certain instances, it does not consistently answer all questions accurately. This paper explores the current limitations of LLMs in navigating complex mathematical problem-solving. Through case analysis, we shed light on the gaps in LLM capabilities within mathematics, thereby setting the stage for future research and advancements in AI-driven mathematical reasoning. We make our code and findings publicly available for research: \url{https://github.com/gipplab/LLM-Investig-MathStackExchange}
The surge of interest in data augmentation within the realm of NLP has been driven by the need to address challenges posed by hate speech domains, the dynamic nature of social media vocabulary, and the demands for large-scale neural networks requiring extensive training data. However, the prevalent use of lexical substitution in data augmentation has raised concerns, as it may inadvertently alter the intended meaning, thereby impacting the efficacy of supervised machine learning models. In pursuit of suitable data augmentation methods, this study explores both established legacy approaches and contemporary practices such as Large Language Models (LLM), including GPT in Hate Speech detection. Additionally, we propose an optimized utilization of BERT-based encoder models with contextual cosine similarity filtration, exposing significant limitations in prior synonym substitution methods. Our comparative analysis encompasses five popular augmentation techniques: WordNet and Fast-Text synonym replacement, Back-translation, BERT-mask contextual augmentation, and LLM. Our analysis across five benchmarked datasets revealed that while traditional methods like back-translation show low label alteration rates (0.3-1.5%), and BERT-based contextual synonym replacement offers sentence diversity but at the cost of higher label alteration rates (over 6%). Our proposed BERT-based contextual cosine similarity filtration markedly reduced label alteration to just 0.05%, demonstrating its efficacy in 0.7% higher F1 performance. However, augmenting data with GPT-3 not only avoided overfitting with up to sevenfold data increase but also improved embedding space coverage by 15% and classification F1 score by 1.4% over traditional methods, and by 0.8% over our method.
With the rapid development and widespread application of Large Language Models (LLMs), the use of Machine-Generated Text (MGT) has become increasingly common, bringing with it potential risks, especially in terms of quality and integrity in fields like news, education, and science. Current research mainly focuses on purely MGT detection without adequately addressing mixed scenarios, including AI-revised Human-Written Text (HWT) or human-revised MGT. To tackle this challenge, we define mixtext, a form of mixed text involving both AI and human-generated content. Then, we introduce MixSet, the first dataset dedicated to studying these mixtext scenarios. Leveraging MixSet, we executed comprehensive experiments to assess the efficacy of prevalent MGT detectors in handling mixtext situations, evaluating their performance in terms of effectiveness, robustness, and generalization. Our findings reveal that existing detectors struggle to identify mixtext, particularly in dealing with subtle modifications and style adaptability. This research underscores the urgent need for more fine-grain detectors tailored for mixtext, offering valuable insights for future research. Code and Models are available at https://github.com/Dongping-Chen/MixSet.
This study introduces GPTA, a Large Language Model assistance training framework, that enhances the training of downstream task models via prefix prompt. By minimizing data exposure to LLM, the framework addresses the security and legal challenges of applying LLM in downstream task model training. GPTA utilizes a new synergistic training approach, optimizing the downstream models with parameter gradients and LLMs with the novel ``dialogue gradient''. The framework not only demonstrates significant improvements in model performance across six NLP benchmark datasets, but also reduces overfitting in low-resource scenarios effectively. The detailed analyses further validate that our pioneer framework provides a cost-efficient and adaptive method for downstream task model training with LLM support.
This position paper argues that large language models (LLMs) constitute promising yet underutilized academic reading companions capable of enhancing learning. We detail an exploratory study examining Claude from Anthropic, an LLM-based interactive assistant that helps students comprehend complex qualitative literature content. The study compares quantitative survey data and qualitative interviews assessing outcomes between a control group and an experimental group leveraging Claude over a semester across two graduate courses. Initial findings demonstrate tangible improvements in reading comprehension and engagement among participants using the AI agent versus unsupported independent study. However, there is potential for overreliance and ethical considerations that warrant continued investigation. By documenting an early integration of an LLM reading companion into an educational context, this work contributes pragmatic insights to guide development of synthetic personae supporting learning. Broader impacts compel policy and industry actions to uphold responsible design in order to maximize benefits of AI integration while prioritizing student wellbeing.
With the ubiquitous use of modern large language models (LLMs) across industries, the inference serving for these models is ever expanding. Given the high compute and memory requirements of modern LLMs, more and more top-of-the-line GPUs are being deployed to serve these models. Energy availability has come to the forefront as the biggest challenge for data center expansion to serve these models. In this paper, we present the trade-offs brought up by making energy efficiency the primary goal of LLM serving under performance SLOs. We show that depending on the inputs, the model, and the service-level agreements, there are several knobs available to the LLM inference provider to use for being energy efficient. We characterize the impact of these knobs on the latency, throughput, as well as the energy. By exploring these trade-offs, we offer valuable insights into optimizing energy usage without compromising on performance, thereby paving the way for sustainable and cost-effective LLM deployment in data center environments.
We consider the problem of aligning a large language model (LLM) to model the preferences of a human population. Modeling the beliefs, preferences, and behaviors of a specific population can be useful for a variety of different applications, such as conducting simulated focus groups for new products, conducting virtual surveys, and testing behavioral interventions, especially for interventions that are expensive, impractical, or unethical. Existing work has had mixed success using LLMs to accurately model human behavior in different contexts. We benchmark and evaluate two well-known fine-tuning approaches and evaluate the resulting populations on their ability to match the preferences of real human respondents on a survey of preferences for battery electric vehicles (BEVs). We evaluate our models against their ability to match population-wide statistics as well as their ability to match individual responses, and we investigate the role of temperature in controlling the trade-offs between these two. Additionally, we propose and evaluate a novel loss term to improve model performance on responses that require a numeric response.
The demand for inference on extremely large scale LLMs has seen enormous growth in the recent months. It made evident the colossal shortage of dedicated hardware capable of efficient and fast processing of the involved compute and memory movement. The problem is aggravated by the exploding raise in the lengths of the sequences being processed, since those require efficient on-chip storage of the KV-cache of size proportional to the sequence length. To make the required compute feasible and fit the involved data into available memory, numerous quantization techniques have been proposed that allow accurate quantization for both weights and activations. One of the main recent breakthroughs in this direction was introduction of the family of Block Floating Point (BFP) formats characterized by a block of mantissas with a shared scale factor. These enable memory- power-, and compute- efficient hardware support of the tensor operations and provide extremely good quantization accuracy. The main issues preventing widespread application of block formats is caused by the presence of outliers in weights and activations since those affect the accuracy of the other values in the same block. In this paper, we focus on the most critical problem of limited KV-cache storage. We propose a novel approach enabling usage of low precision BFP formats without compromising the resulting model accuracy. We exploit the common channel-wise patterns exhibited by the outliers to rearrange them in such a way, that their quantization quality is significantly improved. The methodology yields 2x savings in the memory footprint without significant degradation of the model's accuracy. Importantly, the rearrangement of channels happens at the compile time and thus has no impact on the inference latency.
Large language models (LLMs) have achieved remarkable breakthroughs in new dialogue capabilities by leveraging instruction tuning, which refreshes human impressions of dialogue systems. The long-standing goal of dialogue systems is to be human-like enough to establish long-term connections with users. Therefore, there has been an urgent need to evaluate LLMs as human-like dialogue systems. In this paper, we propose DialogBench, a dialogue evaluation benchmark that contains 12 dialogue tasks to probe the capabilities of LLMs as human-like dialogue systems should have. Specifically, we prompt GPT-4 to generate evaluation instances for each task. We first design the basic prompt based on widely used design principles and further mitigate the existing biases to generate higher-quality evaluation instances. Our extensive tests on English and Chinese DialogBench of 26 LLMs show that instruction tuning improves the human likeness of LLMs to a certain extent, but most LLMs still have much room for improvement as human-like dialogue systems. Interestingly, results also show that the positioning of assistant AI can make instruction tuning weaken the human emotional perception of LLMs and their mastery of information about human daily life.
Modern transformer-based Large Language Models (LLMs) are constructed with a series of decoder blocks. Each block comprises three key components: (1) QKV generation, (2) multi-head attention, and (3) feed-forward networks. In batched processing, QKV generation and feed-forward networks involve compute-intensive matrix-matrix multiplications (GEMM), while multi-head attention requires bandwidth-heavy matrix-vector multiplications (GEMV). Machine learning accelerators like TPUs or NPUs are proficient in handling GEMM but are less efficient for GEMV computations. Conversely, Processing-in-Memory (PIM) technology is tailored for efficient GEMV computation, while it lacks the computational power to handle GEMM effectively. Inspired by this insight, we propose NeuPIMs, a heterogeneous acceleration system that jointly exploits a conventional GEMM-focused NPU and GEMV-optimized PIM devices. The main challenge in efficiently integrating NPU and PIM lies in enabling concurrent operations on both platforms, each addressing a specific kernel type. First, existing PIMs typically operate in a "blocked" mode, allowing only either NPU or PIM to be active at any given time. Second, the inherent dependencies between GEMM and GEMV in LLMs restrict their parallel processing. To tackle these challenges, NeuPIMs is equipped with dual row buffers in each bank, facilitating the simultaneous management of memory read/write operations and PIM commands. Further, NeuPIMs employs a runtime sub-batch interleaving technique to maximize concurrent execution, leveraging batch parallelism to allow two independent sub-batches to be pipelined within a single NeuPIMs device. Our evaluation demonstrates that compared to GPU-only, NPU-only, and a na\"ive NPU+PIM integrated acceleration approaches, NeuPIMs achieves 3$\times$, 2.4$\times$ and 1.6$\times$ throughput improvement, respectively.
Large language models (LLMs) recently exhibited remarkable reasoning capabilities on solving math problems. To further improve their reasoning capabilities, this work explores whether LLMs can LEarn from MistAkes (LEMA), akin to the human learning process. Consider a human student who failed to solve a math problem, he will learn from what mistake he has made and how to correct it. Mimicking this error-driven learning process, LEMA incorporates mistake-correction data pairs during fine-tuning LLMs. Specifically, we first collect inaccurate reasoning paths from various LLMs, and then employ GPT-4 as a ''corrector'' to identify the mistake step, explain the reason for the mistake, correct the mistake and generate the final answer. In addition, we apply a correction-centric evolution strategy that effectively expands the question set for generating correction data. Experiments across various LLMs and reasoning tasks show that LEMA effectively improves CoT-alone fine-tuning. Our further ablations shed light on the non-homogeneous effectiveness between CoT data and correction data. These results suggest a significant potential for LLMs to improve through learning from their mistakes. Our code, models and prompts are publicly available at https://github.com/microsoft/LEMA.
As personalized recommendation systems become vital in the age of information overload, traditional methods relying solely on historical user interactions often fail to fully capture the multifaceted nature of human interests. To enable more human-centric modeling of user preferences, this work proposes a novel explainable recommendation framework, i.e., LLMHG, synergizing the reasoning capabilities of large language models (LLMs) and the structural advantages of hypergraph neural networks. By effectively profiling and interpreting the nuances of individual user interests, our framework pioneers enhancements to recommendation systems with increased explainability. We validate that explicitly accounting for the intricacies of human preferences allows our human-centric and explainable LLMHG approach to consistently outperform conventional models across diverse real-world datasets. The proposed plug-and-play enhancement framework delivers immediate gains in recommendation performance while offering a pathway to apply advanced LLMs for better capturing the complexity of human interests across machine learning applications.
While large language models (LLMs) excel in a simulated world of texts, they struggle to interact with the more realistic world without perceptions of other modalities such as visual or audio signals. Although vision-language models (VLMs) integrate LLM modules (1) aligned with static image features, and (2) may possess prior knowledge of world dynamics (as demonstrated in the text world), they have not been trained in an embodied visual world and thus cannot align with its dynamics. On the other hand, training an embodied agent in a noisy visual world without expert guidance is often challenging and inefficient. In this paper, we train a VLM agent living in a visual world using an LLM agent excelling in a parallel text world. Specifically, we distill LLM's reflection outcomes (improved actions by analyzing mistakes) in a text world's tasks to finetune the VLM on the same tasks of the visual world, resulting in an Embodied Multi-Modal Agent (EMMA) quickly adapting to the visual world dynamics. Such cross-modality imitation learning between the two parallel worlds is achieved by a novel DAgger-DPO algorithm, enabling EMMA to generalize to a broad scope of new tasks without any further guidance from the LLM expert. Extensive evaluations on the ALFWorld benchmark's diverse tasks highlight EMMA's superior performance to SOTA VLM-based agents, e.g., 20%-70% improvement in the success rate.
Open-source pre-trained Large Language Models (LLMs) exhibit strong language understanding and generation capabilities, making them highly successful in a variety of tasks. However, when used as agents for dealing with complex problems in the real world, their performance is far inferior to large commercial models such as ChatGPT and GPT-4. As intelligent agents, LLMs need to have the capabilities of task planning, long-term memory, and the ability to leverage external tools to achieve satisfactory performance. Various methods have been proposed to enhance the agent capabilities of LLMs. On the one hand, methods involve constructing agent-specific data and fine-tuning the models. On the other hand, some methods focus on designing prompts that effectively activate the reasoning abilities of the LLMs. We explore both strategies on the 7B and 13B models. We propose a comprehensive method for constructing agent-specific data using GPT-4. Through supervised fine-tuning with constructed data, we find that for these models with a relatively small number of parameters, supervised fine-tuning can significantly reduce hallucination outputs and formatting errors in agent tasks. Furthermore, techniques such as multi-path reasoning and task decomposition can effectively decrease problem complexity and enhance the performance of LLMs as agents. We evaluate our method on five agent tasks of AgentBench and achieve satisfactory results.
The recent success of Large Language Models (LLMs) has garnered significant attention in both academia and industry. Prior research on LLMs has primarily focused on enhancing or leveraging their generalization capabilities in zero- and few-shot settings. However, there has been limited investigation into effectively fine-tuning LLMs for a specific natural language understanding task in supervised settings. In this study, we conduct an experimental analysis by fine-tuning LLMs for the task of Chinese short text matching. We explore various factors that influence performance when fine-tuning LLMs, including task modeling methods, prompt formats, and output formats.
"I'm categorizing LLM as a productivity tool": Examining ethics of LLM use in HCI research practices
Large language models are increasingly applied in real-world scenarios, including research and education. These models, however, come with well-known ethical issues, which may manifest in unexpected ways in human-computer interaction research due to the extensive engagement with human subjects. This paper reports on research practices related to LLM use, drawing on 16 semi-structured interviews and a survey conducted with 50 HCI researchers. We discuss the ways in which LLMs are already being utilized throughout the entire HCI research pipeline, from ideation to system development and paper writing. While researchers described nuanced understandings of ethical issues, they were rarely or only partially able to identify and address those ethical concerns in their own projects. This lack of action and reliance on workarounds was explained through the perceived lack of control and distributed responsibility in the LLM supply chain, the conditional nature of engaging with ethics, and competing priorities. Finally, we reflect on the implications of our findings and present opportunities to shape emerging norms of engaging with large language models in HCI research.
Most studies on machine learning in sensing systems focus on low-level perception tasks that process raw sensory data within a short time window. However, many practical applications, such as human routine modeling and occupancy tracking, require high-level reasoning abilities to comprehend concepts and make inferences based on long-term sensor traces. Existing machine learning-based approaches for handling such complex tasks struggle to generalize due to the limited training samples and the high dimensionality of sensor traces, necessitating the integration of human knowledge for designing first-principle models or logic reasoning methods. We pose a fundamental question: Can we harness the reasoning capabilities and world knowledge of Large Language Models (LLMs) to recognize complex events from long-term spatiotemporal sensor traces? To answer this question, we design an effective prompting framework for LLMs on high-level reasoning tasks, which can handle traces from the raw sensor data as well as the low-level perception results. We also design two strategies to enhance performance with long sensor traces, including summarization before reasoning and selective inclusion of historical traces. Our framework can be implemented in an edge-cloud setup, running small LLMs on the edge for data summarization and performing high-level reasoning on the cloud for privacy preservation. The results show that LLMSense can achieve over 80\% accuracy on two high-level reasoning tasks such as dementia diagnosis with behavior traces and occupancy tracking with environmental sensor traces. This paper provides a few insights and guidelines for leveraging LLM for high-level reasoning on sensor traces and highlights several directions for future work.
As the range of applications for Large Language Models (LLMs) continues to grow, the demand for effective serving solutions becomes increasingly critical. Despite the versatility of LLMs, no single model can optimally address all tasks and applications, particularly when balancing performance with cost. This limitation has led to the development of LLM routing systems, which combine the strengths of various models to overcome the constraints of individual LLMs. Yet, the absence of a standardized benchmark for evaluating the performance of LLM routers hinders progress in this area. To bridge this gap, we present RouterBench, a novel evaluation framework designed to systematically assess the efficacy of LLM routing systems, along with a comprehensive dataset comprising over 405k inference outcomes from representative LLMs to support the development of routing strategies. We further propose a theoretical framework for LLM routing, and deliver a comparative analysis of various routing approaches through RouterBench, highlighting their potentials and limitations within our evaluation framework. This work not only formalizes and advances the development of LLM routing systems but also sets a standard for their assessment, paving the way for more accessible and economically viable LLM deployments. The code and data are available at https://github.com/withmartian/routerbench.
Understanding data visualizations like charts and plots requires reasoning about both visual elements and numerics. Although strong in extractive questions, current chart visual question answering (chart VQA) models suffer on complex reasoning questions. In this work, we address the lack of reasoning ability by data augmentation. We leverage Large Language Models (LLMs), which have shown to have strong reasoning ability, as an automatic data annotator that generates question-answer annotations for chart images. The key innovation in our method lies in the Synthesize Step-by-Step strategy: our LLM-based data generator learns to decompose the complex question into step-by-step sub-questions (rationales), which are then used to derive the final answer using external tools, i.e. Python. This step-wise generation procedure is trained on synthetic data generated using a template-based QA generation pipeline. Experimental results highlight the significance of the proposed step-by-step generation. By training with the LLM-augmented data (LAMENDA), we significantly enhance the chart VQA models, achieving the state-of-the-art accuracy on the ChartQA and PlotQA datasets. In particular, our approach improves the accuracy of the previous state-of-the-art approach from 38% to 54% on the human-written questions in the ChartQA dataset, which needs strong reasoning. We hope our work underscores the potential of synthetic data and encourages further exploration of data augmentation using LLMs for reasoning-heavy tasks.
The rapid proliferation of large language models (LLMs) such as GPT-4 and Gemini underscores the intense demand for resources during their training processes, posing significant challenges due to substantial computational and environmental costs. To alleviate this issue, we propose checkpoint merging in pretraining LLM. This method utilizes LLM checkpoints with shared training trajectories, and is rooted in an extensive search space exploration for the best merging weight via Bayesian optimization. Through various experiments, we demonstrate that: (1) Our proposed methodology exhibits the capacity to augment pretraining, presenting an opportunity akin to obtaining substantial benefits at minimal cost; (2) Our proposed methodology, despite requiring a given held-out dataset, still demonstrates robust generalization capabilities across diverse domains, a pivotal aspect in pretraining.
With the rise of large language models (LLMs), recent works have leveraged LLMs to improve the performance of click-through rate (CTR) prediction. However, we argue that a critical obstacle remains in deploying LLMs for practical use: the efficiency of LLMs when processing long textual user behaviors. As user sequences grow longer, the current efficiency of LLMs is inadequate for training on billions of users and items. To break through the efficiency barrier of LLMs, we propose Behavior Aggregated Hierarchical Encoding (BAHE) to enhance the efficiency of LLM-based CTR modeling. Specifically, BAHE proposes a novel hierarchical architecture that decouples the encoding of user behaviors from inter-behavior interactions. Firstly, to prevent computational redundancy from repeated encoding of identical user behaviors, BAHE employs the LLM's pre-trained shallow layers to extract embeddings of the most granular, atomic user behaviors from extensive user sequences and stores them in the offline database. Subsequently, the deeper, trainable layers of the LLM facilitate intricate inter-behavior interactions, thereby generating comprehensive user embeddings. This separation allows the learning of high-level user representations to be independent of low-level behavior encoding, significantly reducing computational complexity. Finally, these refined user embeddings, in conjunction with correspondingly processed item embeddings, are incorporated into the CTR model to compute the CTR scores. Extensive experimental results show that BAHE reduces training time and memory by five times for CTR models using LLMs, especially with longer user sequences. BAHE has been deployed in a real-world system, allowing for daily updates of 50 million CTR data on 8 A100 GPUs, making LLMs practical for industrial CTR prediction.
Transformer-based large language models (LLMs) have demonstrated significant potential in addressing logic problems. capitalizing on the great capabilities of LLMs for code-related activities, several frameworks leveraging logical solvers for logic reasoning have been proposed recently. While existing research predominantly focuses on viewing LLMs as natural language logic solvers or translators, their roles as logic code interpreters and executors have received limited attention. This study delves into a novel aspect, namely logic code simulation, which forces LLMs to emulate logical solvers in predicting the results of logical programs. To further investigate this novel task, we formulate our three research questions: Can LLMs efficiently simulate the outputs of logic codes? What strength arises along with logic code simulation? And what pitfalls? To address these inquiries, we curate three novel datasets tailored for the logic code simulation task and undertake thorough experiments to establish the baseline performance of LLMs in code simulation. Subsequently, we introduce a pioneering LLM-based code simulation technique, Dual Chains of Logic (DCoL). This technique advocates a dual-path thinking approach for LLMs, which has demonstrated state-of-the-art performance compared to other LLM prompt strategies, achieving a notable improvement in accuracy by 7.06% with GPT-4-Turbo.
LLMs have become the go-to choice for code generation tasks, with an exponential increase in the training, development, and usage of LLMs specifically for code generation. To evaluate the ability of LLMs on code, both academic and industry practitioners rely on popular handcrafted benchmarks. However, prior benchmarks contain only a very limited set of problems, both in quantity and variety. Further, due to popularity and age, many benchmarks are prone to data leakage where example solutions can be readily found on the web and thus potentially in training data. Such limitations inevitably lead us to inquire: Is the leaderboard performance on existing benchmarks reliable and comprehensive enough to measure the program synthesis ability of LLMs? To address this, we introduce EvoEval -- a program synthesis benchmark suite created by evolving existing benchmarks into different targeted domains for a comprehensive evaluation of LLM coding abilities. Our study on 51 LLMs shows that compared to the high performance obtained on standard benchmarks like HumanEval, there is a significant drop in performance (on average 39.4%) when using EvoEval. Additionally, the decrease in performance can range from 19.6% to 47.7%, leading to drastic ranking changes amongst LLMs and showing potential overfitting of existing benchmarks. Furthermore, we showcase various insights, including the brittleness of instruction-following models when encountering rewording or subtle changes as well as the importance of learning problem composition and decomposition. EvoEval not only provides comprehensive benchmarks, but can be used to further evolve arbitrary problems to keep up with advances and the ever-changing landscape of LLMs for code. We have open-sourced our benchmarks, tools, and complete LLM generations at https://github.com/evo-eval/evoeval
Large Language Models (LLMs) are now commonplace in conversation applications. However, their risks of misuse for generating harmful responses have raised serious societal concerns and spurred recent research on LLM conversation safety. Therefore, in this survey, we provide a comprehensive overview of recent studies, covering three critical aspects of LLM conversation safety: attacks, defenses, and evaluations. Our goal is to provide a structured summary that enhances understanding of LLM conversation safety and encourages further investigation into this important subject. For easy reference, we have categorized all the studies mentioned in this survey according to our taxonomy, available at: https://github.com/niconi19/LLM-conversation-safety.
Verifiable generation aims to let the large language model (LLM) generate text with supporting documents, which enables the user to flexibly verify the answer and makes the LLM's output more reliable. Retrieval plays a crucial role in verifiable generation. Specifically, the retrieved documents not only supplement knowledge to help the LLM generate correct answers, but also serve as supporting evidence for the user to verify the LLM's output. However, the widely used retrievers become the bottleneck of the entire pipeline and limit the overall performance. Their capabilities are usually inferior to LLMs since they often have much fewer parameters than the large language model and have not been demonstrated to scale well to the size of LLMs. If the retriever does not correctly find the supporting documents, the LLM can not generate the correct and verifiable answer, which overshadows the LLM's remarkable abilities. To address these limitations, we propose \LLatrieval (Large Language Model Verified Retrieval), where the LLM updates the retrieval result until it verifies that the retrieved documents can sufficiently support answering the question. Thus, the LLM can iteratively provide feedback to retrieval and facilitate the retrieval result to fully support verifiable generation. Experiments show that LLatrieval significantly outperforms extensive baselines and achieves state-of-the-art results.
To address intricate real-world tasks, there has been a rising interest in tool utilization in applications of large language models (LLMs). To develop LLM-based agents, it usually requires LLMs to understand many tool functions from different tool documentation. But these documentations could be diverse, redundant or incomplete, which immensely affects the capability of LLMs in using tools. To solve this, we introduce EASYTOOL, a framework transforming diverse and lengthy tool documentation into a unified and concise tool instruction for easier tool usage. EasyTool purifies essential information from extensive tool documentation of different sources, and elaborates a unified interface (i.e., tool instruction) to offer standardized tool descriptions and functionalities for LLM-based agents. Extensive experiments on multiple different tasks demonstrate that EasyTool can significantly reduce token consumption and improve the performance of tool utilization in real-world scenarios. Our code will be available at \url{https://github.com/microsoft/JARVIS/} in the future.
Recent explorations with commercial Large Language Models (LLMs) have shown that non-expert users can jailbreak LLMs by simply manipulating their prompts; resulting in degenerate output behavior, privacy and security breaches, offensive outputs, and violations of content regulator policies. Limited studies have been conducted to formalize and analyze these attacks and their mitigations. We bridge this gap by proposing a formalism and a taxonomy of known (and possible) jailbreaks. We survey existing jailbreak methods and their effectiveness on open-source and commercial LLMs (such as GPT-based models, OPT, BLOOM, and FLAN-T5-XXL). We further discuss the challenges of jailbreak detection in terms of their effectiveness against known attacks. For further analysis, we release a dataset of model outputs across 3700 jailbreak prompts over 4 tasks.
Interest is increasing among political scientists in leveraging the extensive information available in images. However, the challenge of interpreting these images lies in the need for specialized knowledge in computer vision and access to specialized hardware. As a result, image analysis has been limited to a relatively small group within the political science community. This landscape could potentially change thanks to the rise of large language models (LLMs). This paper aims to raise awareness of the feasibility of using Gemini for image content analysis. A retrospective analysis was conducted on a corpus of 688 images. Content reports were elicited from Gemini for each image and then manually evaluated by the authors. We find that Gemini is highly accurate in performing object detection, which is arguably the most common and fundamental task in image analysis for political scientists. Equally important, we show that it is easy to implement as the entire command consists of a single prompt in natural language; it is fast to run and should meet the time budget of most researchers; and it is free to use and does not require any specialized hardware. In addition, we illustrate how political scientists can leverage Gemini for other image understanding tasks, including face identification, sentiment analysis, and caption generation. Our findings suggest that Gemini and other similar LLMs have the potential to drastically stimulate and accelerate image research in political science and social sciences more broadly.
Efficient and accurate information extraction from scientific papers is significant in the rapidly developing human-computer interaction research in the literature review process. Our paper introduces and analyses a new information retrieval system using state-of-the-art Large Language Models (LLMs) in combination with structured text analysis techniques to extract experimental data from HCI literature, emphasizing key elements. Then We analyze the challenges and risks of using LLMs in the world of research. We performed a comprehensive analysis on our conducted dataset, which contained the specified information of 300 CHI 2020-2022 papers, to evaluate the performance of the two large language models, GPT-3.5 (text-davinci-003) and Llama-2-70b, paired with structured text analysis techniques. The GPT-3.5 model gains an accuracy of 58\% and a mean absolute error of 7.00. In contrast, the Llama2 model indicates an accuracy of 56\% with a mean absolute error of 7.63. The ability to answer questions was also included in the system in order to work with streamlined data. By evaluating the risks and opportunities presented by LLMs, our work contributes to the ongoing dialogue on establishing methodological validity and ethical guidelines for LLM use in HCI data work.
Large language models (LLM), such as Google's Minerva and OpenAI's GPT families, are becoming increasingly capable of solving mathematical quantitative reasoning problems. However, they still make unjustified logical and computational errors in their reasoning steps and answers. In this paper, we leverage the fact that if the training corpus of LLMs contained sufficiently many examples of formal mathematics (e.g. in Isabelle, a formal theorem proving environment), they can be prompted to translate i.e. autoformalize informal mathematical statements into formal Isabelle code -- which can be verified automatically for internal consistency. This provides a mechanism to automatically reject solutions whose formalized versions are inconsistent within themselves or with the formalized problem statement. We evaluate our method on GSM8K, MATH and MultiArith datasets and demonstrate that our approach provides a consistently better heuristic than vanilla majority voting -- the previously best method to identify correct answers, by more than 12% on GSM8K. In our experiments it improves results consistently across all datasets and LLM model sizes. The code can be found at https://github.com/jinpz/dtv.
Computational notebooks became indispensable tools for research-related development, offering unprecedented interactivity and flexibility in the development process. However, these benefits come at the cost of reproducibility and an increased potential for bugs. There are many tools for bug fixing; however, they are generally targeted at the classical linear code. With the rise of code-fluent Large Language Models, a new stream of smart bug-fixing tools has emerged. However, the applicability of those tools is still problematic for non-linear computational notebooks. In this paper, we propose a potential solution for resolving errors in computational notebooks via an iterative LLM-based agent. We discuss the questions raised by this approach and share a novel dataset of computational notebooks containing bugs to facilitate the research of the proposed approach.
Wireless spectrum regulation is a complex and demanding process due to the rapid pace of technological progress, increasing demand for spectrum, and a multitude of stakeholders with potentially conflicting interests, alongside significant economic implications. To navigate this, regulators must engage effectively with all parties, keep pace with global technology trends, conduct technical evaluations, issue licenses in a timely manner, and comply with various legal and policy frameworks. In light of these challenges, this paper demonstrates example applications of Large Language Models (LLMs) to expedite spectrum regulatory processes. We explore various roles that LLMs can play in this context while identifying some of the challenges to address. The paper also offers practical case studies and insights, with appropriate experiments, highlighting the transformative potential of LLMs in spectrum management.
In this paper, we present TWOLAR: a two-stage pipeline for passage reranking based on the distillation of knowledge from Large Language Models (LLM). TWOLAR introduces a new scoring strategy and a distillation process consisting in the creation of a novel and diverse training dataset. The dataset consists of 20K queries, each associated with a set of documents retrieved via four distinct retrieval methods to ensure diversity, and then reranked by exploiting the zero-shot reranking capabilities of an LLM. Our ablation studies demonstrate the contribution of each new component we introduced. Our experimental results show that TWOLAR significantly enhances the document reranking ability of the underlying model, matching and in some cases even outperforming state-of-the-art models with three orders of magnitude more parameters on the TREC-DL test sets and the zero-shot evaluation benchmark BEIR. To facilitate future work we release our data set, finetuned models, and code.
Integrating LLM and reinforcement learning (RL) agent effectively to achieve complementary performance is critical in high stake tasks like cybersecurity operations. In this study, we introduce SecurityBot, a LLM agent mentored by pre-trained RL agents, to support cybersecurity operations. In particularly, the LLM agent is supported with a profile module to generated behavior guidelines, a memory module to accumulate local experiences, a reflection module to re-evaluate choices, and an action module to reduce action space. Additionally, it adopts the collaboration mechanism to take suggestions from pre-trained RL agents, including a cursor for dynamic suggestion taken, an aggregator for multiple mentors' suggestions ranking and a caller for proactive suggestion asking. Building on the CybORG experiment framework, our experiences show that SecurityBot demonstrates significant performance improvement compared with LLM or RL standalone, achieving the complementary performance in the cybersecurity games.
Attention based Large Language Models (LLMs) are the state-of-the-art in natural language processing (NLP). The two most common architectures are encoders such as BERT, and decoders like the GPT models. Despite the success of encoder models, on which we focus in this work, they also bear several risks, including issues with bias or their susceptibility for adversarial attacks, signifying the necessity for explainable AI to detect such issues. While there does exist various local explainability methods focusing on the prediction of single inputs, global methods based on dimensionality reduction for classification inspection, which have emerged in other domains and that go further than just using t-SNE in the embedding space, are not widely spread in NLP. To reduce this gap, we investigate the application of DeepView, a method for visualizing a part of the decision function together with a data set in two dimensions, to the NLP domain. While in previous work, DeepView has been used to inspect deep image classification models, we demonstrate how to apply it to BERT-based NLP classifiers and investigate its usability in this domain, including settings with adversarially perturbed input samples and pre-trained, fine-tuned, and multi-task models.
Safety lies at the core of the development of Large Language Models (LLMs). There is ample work on aligning LLMs with human ethics and preferences, including data filtering in pretraining, supervised fine-tuning, reinforcement learning from human feedback, and red teaming, etc. In this study, we discover that chat in cipher can bypass the safety alignment techniques of LLMs, which are mainly conducted in natural languages. We propose a novel framework CipherChat to systematically examine the generalizability of safety alignment to non-natural languages -- ciphers. CipherChat enables humans to chat with LLMs through cipher prompts topped with system role descriptions and few-shot enciphered demonstrations. We use CipherChat to assess state-of-the-art LLMs, including ChatGPT and GPT-4 for different representative human ciphers across 11 safety domains in both English and Chinese. Experimental results show that certain ciphers succeed almost 100% of the time to bypass the safety alignment of GPT-4 in several safety domains, demonstrating the necessity of developing safety alignment for non-natural languages. Notably, we identify that LLMs seem to have a ''secret cipher'', and propose a novel SelfCipher that uses only role play and several demonstrations in natural language to evoke this capability. SelfCipher surprisingly outperforms existing human ciphers in almost all cases. Our code and data will be released at https://github.com/RobustNLP/CipherChat.
Knowledge concept tagging for questions plays a crucial role in contemporary intelligent educational applications, including learning progress diagnosis, practice question recommendations, and course content organization. Traditionally, these annotations have been conducted manually with help from pedagogical experts, as the task requires not only a strong semantic understanding of both question stems and knowledge definitions but also deep insights into connecting question-solving logic with corresponding knowledge concepts. In this paper, we explore automating the tagging task using Large Language Models (LLMs), in response to the inability of prior manual methods to meet the rapidly growing demand for concept tagging in questions posed by advanced educational applications. Moreover, the zero/few-shot learning capability of LLMs makes them well-suited for application in educational scenarios, which often face challenges in collecting large-scale, expertise-annotated datasets. By conducting extensive experiments with a variety of representative LLMs, we demonstrate that LLMs are a promising tool for concept tagging in math questions. Furthermore, through case studies examining the results from different LLMs, we draw some empirical conclusions about the key factors for success in applying LLMs to the automatic concept tagging task.
Recent advancements in diffusion models for 2D and 3D content creation have sparked a surge of interest in generating 4D content. However, the scarcity of 3D scene datasets constrains current methodologies to primarily object-centric generation. To overcome this limitation, we present Comp4D, a novel framework for Compositional 4D Generation. Unlike conventional methods that generate a singular 4D representation of the entire scene, Comp4D innovatively constructs each 4D object within the scene separately. Utilizing Large Language Models (LLMs), the framework begins by decomposing an input text prompt into distinct entities and maps out their trajectories. It then constructs the compositional 4D scene by accurately positioning these objects along their designated paths. To refine the scene, our method employs a compositional score distillation technique guided by the pre-defined trajectories, utilizing pre-trained diffusion models across text-to-image, text-to-video, and text-to-3D domains. Extensive experiments demonstrate our outstanding 4D content creation capability compared to prior arts, showcasing superior visual quality, motion fidelity, and enhanced object interactions.
The use of ChatGPT and similar Large Language Model (LLM) tools in scholarly communication and academic publishing has been widely discussed since they became easily accessible to a general audience in late 2022. This study uses keywords known to be disproportionately present in LLM-generated text to provide an overall estimate for the prevalence of LLM-assisted writing in the scholarly literature. For the publishing year 2023, it is found that several of those keywords show a distinctive and disproportionate increase in their prevalence, individually and in combination. It is estimated that at least 60,000 papers (slightly over 1% of all articles) were LLM-assisted, though this number could be extended and refined by analysis of other characteristics of the papers or by identification of further indicative keywords.
In AI-assisted decision-making, humans often passively review AI's suggestion and decide whether to accept or reject it as a whole. In such a paradigm, humans are found to rarely trigger analytical thinking and face difficulties in communicating the nuances of conflicting opinions to the AI when disagreements occur. To tackle this challenge, we propose Human-AI Deliberation, a novel framework to promote human reflection and discussion on conflicting human-AI opinions in decision-making. Based on theories in human deliberation, this framework engages humans and AI in dimension-level opinion elicitation, deliberative discussion, and decision updates. To empower AI with deliberative capabilities, we designed Deliberative AI, which leverages large language models (LLMs) as a bridge between humans and domain-specific models to enable flexible conversational interactions and faithful information provision. An exploratory evaluation on a graduate admissions task shows that Deliberative AI outperforms conventional explainable AI (XAI) assistants in improving humans' appropriate reliance and task performance. Based on a mixed-methods analysis of participant behavior, perception, user experience, and open-ended feedback, we draw implications for future AI-assisted decision tool design.
The increasing prevalence of Cyber-Physical Systems and the Internet of Things (CPS-IoT) applications and Foundation Models are enabling new applications that leverage real-time control of the environment. For example, real-time control of Heating, Ventilation and Air-Conditioning (HVAC) systems can reduce its usage when not needed for the comfort of human occupants, hence reducing energy consumption. Collecting real-time feedback on human preferences in such human-in-the-loop (HITL) systems, however, is difficult in practice. We propose the use of large language models (LLMs) to deal with the challenges of dynamic environments and difficult-to-obtain data in CPS optimization. In this paper, we present a case study that employs LLM agents to mimic the behaviors and thermal preferences of various population groups (e.g. young families, the elderly) in a shopping mall. The aggregated thermal preferences are integrated into an agent-in-the-loop based reinforcement learning algorithm AitL-RL, which employs the LLM as a dynamic simulation of the physical environment to learn how to balance between energy savings and occupant comfort. Our results show that LLMs are capable of simulating complex population movements within large open spaces. Besides, AitL-RL demonstrates superior performance compared to the popular existing policy of set point control, suggesting that adaptive and personalized decision-making is critical for efficient optimization in CPS-IoT applications. Through this case study, we demonstrate the potential of integrating advanced Foundation Models like LLMs into CPS-IoT to enhance system adaptability and efficiency. The project's code can be found on our GitHub repository.
Multimodal Large Language Model (MLLMs) leverages Large Language Models as a cognitive framework for diverse visual-language tasks. Recent efforts have been made to equip MLLMs with visual perceiving and grounding capabilities. However, there still remains a gap in providing fine-grained pixel-level perceptions and extending interactions beyond text-specific inputs. In this work, we propose {\bf{AnyRef}}, a general MLLM model that can generate pixel-wise object perceptions and natural language descriptions from multi-modality references, such as texts, boxes, images, or audio. This innovation empowers users with greater flexibility to engage with the model beyond textual and regional prompts, without modality-specific designs. Through our proposed refocusing mechanism, the generated grounding output is guided to better focus on the referenced object, implicitly incorporating additional pixel-level supervision. This simple modification utilizes attention scores generated during the inference of LLM, eliminating the need for extra computations while exhibiting performance enhancements in both grounding masks and referring expressions. With only publicly available training data, our model achieves state-of-the-art results across multiple benchmarks, including diverse modality referring segmentation and region-level referring expression generation.
Large Language Models (LLMs) have demonstrated remarkable performance across a wide range of natural language processing tasks. However, their enormous parameter size and extremely high requirements for compute power pose challenges for their practical deployment. Recent research has revealed that specific capabilities of LLMs, such as numerical reasoning, can be transferred to smaller models through distillation. Some studies explore the potential of leveraging LLMs to perform table-based reasoning. However, there has been no prior work focusing on table reasoning skills in smaller models specifically tailored for scientific table-to-text generation tasks. In this paper, we propose a novel table-based reasoning distillation approach, with the aim of distilling LLMs into tailored smaller models. Our experimental results have shown that a 220 million parameter model (Flan-T5-base) fine-tuned using distilled data, not only achieves a significant improvement compared to traditionally fine-tuned baselines, but also surpasses specific LLMs on a scientific table-to-text generation dataset. Our code is available at https://github.com/Bernard-Yang/DistillTableCoT.
Large language models (LLMs) are gaining increasing interests to improve clinical efficiency for medical diagnosis, owing to their unprecedented performance in modelling natural language. Ensuring the safe and reliable clinical applications, the evaluation of LLMs indeed becomes critical for better mitigating the potential risks, e.g., hallucinations. However, current evaluation methods heavily rely on labor-intensive human participation to achieve human-preferred judgements. To overcome this challenge, we propose an automatic evaluation paradigm tailored to assess the LLMs' capabilities in delivering clinical services, e.g., disease diagnosis and treatment. The evaluation paradigm contains three basic elements: metric, data, and algorithm. Specifically, inspired by professional clinical practice pathways, we formulate a LLM-specific clinical pathway (LCP) to define the clinical capabilities that a doctor agent should possess. Then, Standardized Patients (SPs) from the medical education are introduced as the guideline for collecting medical data for evaluation, which can well ensure the completeness of the evaluation procedure. Leveraging these steps, we develop a multi-agent framework to simulate the interactive environment between SPs and a doctor agent, which is equipped with a Retrieval-Augmented Evaluation (RAE) to determine whether the behaviors of a doctor agent are in accordance with LCP. The above paradigm can be extended to any similar clinical scenarios to automatically evaluate the LLMs' medical capabilities. Applying such paradigm, we construct an evaluation benchmark in the field of urology, including a LCP, a SPs dataset, and an automated RAE. Extensive experiments are conducted to demonstrate the effectiveness of the proposed approach, providing more insights for LLMs' safe and reliable deployments in clinical practice.
Conversational Recommender System (CRS) interacts with users through natural language to understand their preferences and provide personalized recommendations in real-time. CRS has demonstrated significant potential, prompting researchers to address the development of more realistic and reliable user simulators as a key focus. Recently, the capabilities of Large Language Models (LLMs) have attracted a lot of attention in various fields. Simultaneously, efforts are underway to construct user simulators based on LLMs. While these works showcase innovation, they also come with certain limitations that require attention. In this work, we aim to analyze the limitations of using LLMs in constructing user simulators for CRS, to guide future research. To achieve this goal, we conduct analytical validation on the notable work, iEvaLM. Through multiple experiments on two widely-used datasets in the field of conversational recommendation, we highlight several issues with the current evaluation methods for user simulators based on LLMs: (1) Data leakage, which occurs in conversational history and the user simulator's replies, results in inflated evaluation results. (2) The success of CRS recommendations depends more on the availability and quality of conversational history than on the responses from user simulators. (3) Controlling the output of the user simulator through a single prompt template proves challenging. To overcome these limitations, we propose SimpleUserSim, employing a straightforward strategy to guide the topic toward the target items. Our study validates the ability of CRS models to utilize the interaction information, significantly improving the recommendation results.
Fault Localization (FL) is an essential step during the debugging process. With the strong capabilities of code comprehension, the recent Large Language Models (LLMs) have demonstrated promising performance in diagnosing bugs in the code. Nevertheless, due to LLMs' limited performance in handling long contexts, existing LLM-based fault localization remains on localizing bugs within a small code scope (i.e., a method or a class), which struggles to diagnose bugs for a large code scope (i.e., an entire software system). To address the limitation, this paper presents AgentFL, a multi-agent system based on ChatGPT for automated fault localization. By simulating the behavior of a human developer, AgentFL models the FL task as a three-step process, which involves comprehension, navigation, and confirmation. Within each step, AgentFL hires agents with diversified expertise, each of which utilizes different tools to handle specific tasks. Particularly, we adopt a series of auxiliary strategies such as Test Behavior Tracking, Document-Guided Search, and Multi-Round Dialogue to overcome the challenges in each step. The evaluation on the widely used Defects4J-V1.2.0 benchmark shows that AgentFL can localize 157 out of 395 bugs within Top-1, which outperforms the other LLM-based approaches and exhibits complementarity to the state-of-the-art learning-based techniques. Additionally, we confirm the indispensability of the components in AgentFL with the ablation study and demonstrate the usability of AgentFL through a user study. Finally, the cost analysis shows that AgentFL spends an average of only 0.074 dollars and 97 seconds for a single bug.
In information retrieval, facet identification of a user query is an important task. If a search service can recognize the facets of a user's query, it has the potential to offer users a much broader range of search results. Previous studies can enhance facet prediction by leveraging retrieved documents and related queries obtained through a search engine. However, there are challenges in extending it to other applications when a search engine operates as part of the model. First, search engines are constantly updated. Therefore, additional information may change during training and test, which may reduce performance. The second challenge is that public search engines cannot search for internal documents. Therefore, a separate search system needs to be built to incorporate documents from private domains within the company. We propose two strategies that focus on a framework that can predict facets by taking only queries as input without a search engine. The first strategy is multi-task learning to predict SERP. By leveraging SERP as a target instead of a source, the proposed model deeply understands queries without relying on external modules. The second strategy is to enhance the facets by combining Large Language Model (LLM) and the small model. Overall performance improves when small model and LLM are combined rather than facet generation individually.
In this paper we introduce APL (Arabic Programming Language) that uses Large language models (LLM) as semi-compiler to covert Arabic text code to python code then run the code. Designing a full pipeline from the structure of the APL text then a prompt (using prompt engineering) then running the prodcued python code using PyRunner. This project has a three parts first python library, a playground with simple interface and this research paper.
Large language model (LLM) has marked a pivotal moment in the field of machine learning and deep learning. Recently its capability for query planning has been investigated, including both single-modal and multi-modal queries. However, there is no work on the query optimization capability of LLM. As a critical (or could even be the most important) step that significantly impacts the execution performance of the query plan, such analysis and attempts should not be missed. From another aspect, existing query optimizers are usually rule-based or rule-based + cost-based, i.e., they are dependent on manually created rules to complete the query plan rewrite/transformation. Given the fact that modern optimizers include hundreds to thousands of rules, designing a multi-modal query optimizer following a similar way is significantly time-consuming since we will have to enumerate as many multi-modal optimization rules as possible, which has not been well addressed today. In this paper, we investigate the query optimization ability of LLM and use LLM to design LaPuda, a novel LLM and Policy based multi-modal query optimizer. Instead of enumerating specific and detailed rules, LaPuda only needs a few abstract policies to guide LLM in the optimization, by which much time and human effort are saved. Furthermore, to prevent LLM from making mistakes or negative optimization, we borrow the idea of gradient descent and propose a guided cost descent (GCD) algorithm to perform the optimization, such that the optimization can be kept in the correct direction. In our evaluation, our methods consistently outperform the baselines in most cases. For example, the optimized plans generated by our methods result in 1~3x higher execution speed than those by the baselines.
There is a compelling necessity from enterprises for fine tuning LLMs (Large Language Models) o get them trained on proprietary domain knowledge. The challenge is to imbibe the LLMs with domain specific knowledge using the most optimial resource and cost and in the best possible time. Many enterprises rely on RAG (Retrieval Augmented Generation) which does not need LLMs to be ine-tuned but they are limited by the quality of vector databases and their retrieval capabilities rather than the intrinsic capabilities of the LLMs themselves. In our current work we focus on fine tuning LLaMA, an open source LLM using proprietary documents and code from an enterprise repository and use the fine tuned models to evaluate the quality of responses. As part of this work, we aim to guide beginners on how to start with fine tuning an LLM for documentation and code by making educated guesses on size of GPU required and options that are available for formatting the data. We also propose pre processing recipes for both documentation and code to prepare dataset in different formats. The proposed methods of data preparation for document datasets are forming paragraph chunks, forming question and answer pairs and forming keyword and paragraph chunk pairs. For code dataset we propose forming summary and function pairs. Further, we qualitatively evaluate the results of the models for domain specific queries. Finally, we also propose practical guidelines and recommendations for fine tuning LLMs.
Event temporal relation (TempRel) is a primary subject of the event relation extraction task. However, the inherent ambiguity of TempRel increases the difficulty of the task. With the rise of prompt engineering, it is important to design effective prompt templates and verbalizers to extract relevant knowledge. The traditional manually designed templates struggle to extract precise temporal knowledge. This paper introduces a novel retrieval-augmented TempRel extraction approach, leveraging knowledge retrieved from large language models (LLMs) to enhance prompt templates and verbalizers. Our method capitalizes on the diverse capabilities of various LLMs to generate a wide array of ideas for template and verbalizer design. Our proposed method fully exploits the potential of LLMs for generation tasks and contributes more knowledge to our design. Empirical evaluations across three widely recognized datasets demonstrate the efficacy of our method in improving the performance of event temporal relation extraction tasks.
Instruction-finetuned Large Language Models inherit clear political leanings that have been shown to influence downstream task performance. We expand this line of research beyond the two-party system in the US and audit Llama Chat in the context of EU politics in various settings to analyze the model's political knowledge and its ability to reason in context. We adapt, i.e., further fine-tune, Llama Chat on speeches of individual euro-parties from debates in the European Parliament to reevaluate its political leaning based on the EUandI questionnaire. Llama Chat shows considerable knowledge of national parties' positions and is capable of reasoning in context. The adapted, party-specific, models are substantially re-aligned towards respective positions which we see as a starting point for using chat-based LLMs as data-driven conversational engines to assist research in political science.
Log parsing transforms log messages into structured formats, serving as the prerequisite step for various log analysis tasks. Although a variety of log parsing approaches have been proposed, their performance on complicated log data remains compromised due to the use of human-crafted rules or learning-based models with limited training data. The recent emergence of powerful large language models (LLMs) demonstrates their vast pre-trained knowledge related to code and logging, making it promising to apply LLMs for log parsing. However, their lack of specialized log parsing capabilities currently hinders their accuracy in parsing. Moreover, the inherent inconsistent answers, as well as the substantial overhead, prevent the practical adoption of LLM-based log parsing. To address these challenges, we propose LILAC, the first practical log parsing framework using LLMs with adaptive parsing cache. To facilitate accurate and robust log parsing, LILAC leverages the in-context learning (ICL) capability of the LLM by performing a hierarchical candidate sampling algorithm and selecting high-quality demonstrations. Furthermore, LILAC incorporates a novel component, an adaptive parsing cache, to store and refine the templates generated by the LLM. It helps mitigate LLM's inefficiency issue by enabling rapid retrieval of previously processed log templates. In this process, LILAC adaptively updates the templates within the parsing cache to ensure the consistency of parsed results. The extensive evaluation on public large-scale datasets shows that LILAC outperforms state-of-the-art methods by 69.5% in terms of the average F1 score of template accuracy. In addition, LILAC reduces the query times to LLMs by several orders of magnitude, achieving a comparable efficiency to the fastest baseline.
The jailbreak attack can bypass the safety measures of a Large Language Model (LLM), generating harmful content. This misuse of LLM has led to negative societal consequences. Currently, there are two main approaches to address jailbreak attacks: safety training and safeguards. Safety training focuses on further training LLM to enhance its safety. On the other hand, safeguards involve implementing external models or filters to prevent harmful outputs. However, safety training has constraints in its ability to adapt to new attack types and often leads to a drop in model performance. Safeguards have proven to be of limited help. To tackle these issues, we propose a novel approach called Self-Guard, which combines the strengths of both safety methods. Self-Guard includes two stages. In the first stage, we enhance the model's ability to assess harmful content, and in the second stage, we instruct the model to consistently perform harmful content detection on its own responses. The experiment has demonstrated that Self-Guard is robust against jailbreak attacks. In the bad case analysis, we find that LLM occasionally provides harmless responses to harmful queries. Additionally, we evaluated the general capabilities of the LLM before and after safety training, providing evidence that Self-Guard does not result in the LLM's performance degradation. In sensitivity tests, Self-Guard not only avoids inducing over-sensitivity in LLM but also can even mitigate this issue.
Lexical Simplification (LS) aims to simplify text at the lexical level. Existing methods rely heavily on annotated data, making it challenging to apply in low-resource scenarios. In this paper, we propose a novel LS method without parallel corpora. This method employs an Adversarial Editing System with guidance from a confusion loss and an invariance loss to predict lexical edits in the original sentences. Meanwhile, we introduce an innovative LLM-enhanced loss to enable the distillation of knowledge from Large Language Models (LLMs) into a small-size LS system. From that, complex words within sentences are masked and a Difficulty-aware Filling module is crafted to replace masked positions with simpler words. At last, extensive experimental results and analyses on three benchmark LS datasets demonstrate the effectiveness of our proposed method.
Recently, the remarkable advance of the Large Language Model (LLM) has inspired researchers to transfer its extraordinary reasoning capability to both vision and language data. However, the prevailing approaches primarily regard the visual input as a prompt and focus exclusively on optimizing the text generation process conditioned upon vision content by a frozen LLM. Such an inequitable treatment of vision and language heavily constrains the model's potential. In this paper, we break through this limitation by representing both vision and language in a unified form. Specifically, we introduce a well-designed visual tokenizer to translate the non-linguistic image into a sequence of discrete tokens like a foreign language that LLM can read. The resulting visual tokens encompass high-level semantics worthy of a word and also support dynamic sequence length varying from the image. Coped with this tokenizer, the presented foundation model called LaVIT can handle both image and text indiscriminately under the same generative learning paradigm. This unification empowers LaVIT to serve as an impressive generalist interface to understand and generate multi-modal content simultaneously. Extensive experiments further showcase that it outperforms the existing models by a large margin on massive vision-language tasks. Our code and models are available at https://github.com/jy0205/LaVIT.
With the emergence of numerous Large Language Models (LLM), the usage of such models in various Natural Language Processing (NLP) applications is increasing extensively. Counterspeech generation is one such key task where efforts are made to develop generative models by fine-tuning LLMs with hatespeech - counterspeech pairs, but none of these attempts explores the intrinsic properties of large language models in zero-shot settings. In this work, we present a comprehensive analysis of the performances of four LLMs namely GPT-2, DialoGPT, ChatGPT and FlanT5 in zero-shot settings for counterspeech generation, which is the first of its kind. For GPT-2 and DialoGPT, we further investigate the deviation in performance with respect to the sizes (small, medium, large) of the models. On the other hand, we propose three different prompting strategies for generating different types of counterspeech and analyse the impact of such strategies on the performance of the models. Our analysis shows that there is an improvement in generation quality for two datasets (17%), however the toxicity increase (25%) with increase in model size. Considering type of model, GPT-2 and FlanT5 models are significantly better in terms of counterspeech quality but also have high toxicity as compared to DialoGPT. ChatGPT are much better at generating counter speech than other models across all metrics. In terms of prompting, we find that our proposed strategies help in improving counter speech generation across all the models.
By design, large language models (LLMs) are static general-purpose models, expensive to retrain or update frequently. As they are increasingly adopted for knowledge-intensive tasks, it becomes evident that these design choices lead to failures to generate factual, relevant, and up-to-date knowledge. To this end, we propose Knowledge Card, a modular framework to plug in new factual and relevant knowledge into general-purpose LLMs. We first introduce knowledge cards -- specialized language models trained on corpora from specific domains and sources. Knowledge cards serve as parametric repositories that are selected at inference time to generate background knowledge for the base LLM. We then propose three content selectors to dynamically select and retain information in documents generated by knowledge cards, specifically controlling for relevance, brevity, and factuality of outputs. Finally, we propose two complementary integration approaches to augment the base LLM with the (relevant, factual) knowledge curated from the specialized LMs. Through extensive experiments, we demonstrate that Knowledge Card achieves state-of-the-art performance on six benchmark datasets. Ultimately, Knowledge Card framework enables dynamic synthesis and updates of knowledge from diverse domains. Its modularity will ensure that relevant knowledge can be continuously updated through the collective efforts of the research community.
Teachers' mathematical content knowledge (CK) is of vital importance and need in teacher professional development (PD) programs. Computer-aided asynchronous PD systems are the most recent proposed PD techniques, which aim to help teachers improve their PD equally with fewer concerns about costs and limitations of time or location. However, current automatic CK identification methods, which serve as one of the core techniques of asynchronous PD systems, face challenges such as diversity of user responses, scarcity of high-quality annotated data, and low interpretability of the predictions. To tackle these challenges, we propose a Multi-Agent LLMs-based framework, LLMAgent-CK, to assess the user responses' coverage of identified CK learning goals without human annotations. By taking advantage of multi-agent LLMs in strong generalization ability and human-like discussions, our proposed LLMAgent-CK presents promising CK identifying performance on a real-world mathematical CK dataset MaCKT. Moreover, our case studies further demonstrate the working of the multi-agent framework.
For effective human-robot interaction, robots need to understand, plan, and execute complex, long-horizon tasks described by natural language. Recent advances in large language models (LLMs) have shown promise for translating natural language into robot action sequences for complex tasks. However, existing approaches either translate the natural language directly into robot trajectories or factor the inference process by decomposing language into task sub-goals and relying on a motion planner to execute each sub-goal. When complex environmental and temporal constraints are involved, inference over planning tasks must be performed jointly with motion plans using traditional task-and-motion planning (TAMP) algorithms, making factorization into subgoals untenable. Rather than using LLMs to directly plan task sub-goals, we instead perform few-shot translation from natural language task descriptions to an intermediate task representation that can then be consumed by a TAMP algorithm to jointly solve the task and motion plan. To improve translation, we automatically detect and correct both syntactic and semantic errors via autoregressive re-prompting, resulting in significant improvements in task completion. We show that our approach outperforms several methods using LLMs as planners in complex task domains. See our project website https://yongchao98.github.io/MIT-REALM-AutoTAMP/ for prompts, videos, and code.
In this paper we present the results of an evaluation study of the perfor-mance of LLMs on Technical Language Processing tasks. Humans are often confronted with tasks in which they have to gather information from dispar-ate sources and require making sense of large bodies of text. These tasks can be significantly complex for humans and often require deep study including rereading portions of a text. Towards simplifying the task of gathering in-formation we evaluated LLMs with chat interfaces for their ability to provide answers to standard questions that a human can be expected to answer based on their reading of a body of text. The body of text under study is Title 47 of the United States Code of Federal Regulations (CFR) which describes regula-tions for commercial telecommunications as governed by the Federal Com-munications Commission (FCC). This has been a body of text of interest be-cause our larger research concerns the issue of making sense of information related to Wireless Spectrum Governance and usage in an automated manner to support Dynamic Spectrum Access. The information concerning this wireless spectrum domain is found in many disparate sources, with Title 47 of the CFR being just one of many. Using a range of LLMs and providing the required CFR text as context we were able to quantify the performance of those LLMs on the specific task of answering the questions below.
Large Language Models (LLMs) increasingly support applications in a wide range of domains, some with potential high societal impact such as biomedicine, yet their reliability in realistic use cases is under-researched. In this work we introduce the Reliability AssesMent for Biomedical LLM Assistants (RAmBLA) framework and evaluate whether four state-of-the-art foundation LLMs can serve as reliable assistants in the biomedical domain. We identify prompt robustness, high recall, and a lack of hallucinations as necessary criteria for this use case. We design shortform tasks and tasks requiring LLM freeform responses mimicking real-world user interactions. We evaluate LLM performance using semantic similarity with a ground truth response, through an evaluator LLM.
This paper explores the use of large language models (LLMs) to score and explain short-answer assessments in K-12 science. While existing methods can score more structured math and computer science assessments, they often do not provide explanations for the scores. Our study focuses on employing GPT-4 for automated assessment in middle school Earth Science, combining few-shot and active learning with chain-of-thought reasoning. Using a human-in-the-loop approach, we successfully score and provide meaningful explanations for formative assessment responses. A systematic analysis of our method's pros and cons sheds light on the potential for human-in-the-loop techniques to enhance automated grading for open-ended science assessments.
With the introduction of ChatGPT, Large Language Models (LLMs) have received enormous attention in healthcare. Despite their potential benefits, researchers have underscored various ethical implications. While individual instances have drawn much attention, the debate lacks a systematic overview of practical applications currently researched and ethical issues connected to them. Against this background, this work aims to map the ethical landscape surrounding the current stage of deployment of LLMs in medicine and healthcare. Electronic databases and preprint servers were queried using a comprehensive search strategy. Studies were screened and extracted following a modified rapid review approach. Methodological quality was assessed using a hybrid approach. For 53 records, a meta-aggregative synthesis was performed. Four fields of applications emerged and testify to a vivid exploration phase. Advantages of using LLMs are attributed to their capacity in data analysis, personalized information provisioning, support in decision-making, mitigating information loss and enhancing information accessibility. However, we also identifies recurrent ethical concerns connected to fairness, bias, non-maleficence, transparency, and privacy. A distinctive concern is the tendency to produce harmful misinformation or convincingly but inaccurate content. A recurrent plea for ethical guidance and human oversight is evident. Given the variety of use cases, it is suggested that the ethical guidance debate be reframed to focus on defining what constitutes acceptable human oversight across the spectrum of applications. This involves considering diverse settings, varying potentials for harm, and different acceptable thresholds for performance and certainty in healthcare. In addition, a critical inquiry is necessary to determine the extent to which the current experimental use of LLMs is necessary and justified.
Translation-tailored Large language models (LLMs) exhibit remarkable translation capabilities, even competing with supervised-trained commercial translation systems. However, off-target translation remains an unsolved problem, especially for low-resource languages, hindering us from developing accurate LLMs-based translation models. To mitigate the off-target translation problem and enhance the performance of LLMs on translation, recent works have either designed advanced prompting strategies to highlight the functionality of translation instructions or exploited the in-context learning ability of LLMs by feeding few-shot demonstrations. However, these methods essentially do not improve LLM's ability to follow translation instructions, especially the language direction information. In this work, we design a two-stage fine-tuning algorithm to improve the instruction-following ability (especially the translation direction) of LLMs. Specifically, we first tune LLMs with the maximum likelihood estimation loss on the translation dataset to elicit the basic translation capabilities. In the second stage, we construct instruction-conflicting samples by randomly replacing the translation directions with a wrong one within the instruction, and then introduce an extra unlikelihood loss to learn those samples. Experiments on IWSLT and WMT benchmarks upon the LLaMA model spanning 16 zero-shot directions show that, compared to the competitive baseline -- translation-finetuned LLama, our method could effectively reduce the off-target translation ratio (averagely -53.3\%), thus improving translation quality with average +5.7 SacreBLEU and +16.4 BLEURT. Analysis shows that our method could preserve the model's general task performance on AlpacaEval. Code and models will be released at \url{https://github.com/alphadl/LanguageAware_Tuning}.
Large Language Models (LLMs) have shown great potential in Natural Language Processing (NLP) tasks. However, recent literature reveals that LLMs generate nonfactual responses intermittently, which impedes the LLMs' reliability for further utilization. In this paper, we propose a novel self-detection method to detect which questions that a LLM does not know that are prone to generate nonfactual results. Specifically, we first diversify the textual expressions for a given question and collect the corresponding answers. Then we examine the divergencies between the generated answers to identify the questions that the model may generate falsehoods. All of the above steps can be accomplished by prompting the LLMs themselves without referring to any other external resources. We conduct comprehensive experiments and demonstrate the effectiveness of our method on recently released LLMs, e.g., Vicuna, ChatGPT, and GPT-4.
Already since the 1950s TRIZ shows that patents and the technical contradictions they solve are an important source of inspiration for the development of innovative products. However, TRIZ is a heuristic based on a historic patent analysis and does not make use of the ever-increasing number of latest technological solutions in current patents. Because of the huge number of patents, their length, and, last but not least, their complexity there is a need for modern patent retrieval and patent analysis to go beyond keyword-oriented methods. Recent advances in patent retrieval and analysis mainly focus on dense vectors based on neural AI Transformer language models like Google BERT. They are, for example, used for dense retrieval, question answering or summarization and key concept extraction. A research focus within the methods for patent summarization and key concept extraction are generic inventive concepts respectively TRIZ concepts like problems, solutions, advantage of invention, parameters, and contradictions. Succeeding rule-based approaches, finetuned BERT-like language models for sentence-wise classification represent the state-of-the-art of inventive concept extraction. While they work comparatively well for basic concepts like problems or solutions, contradictions - as a more complex abstraction - remain a challenge for these models. This paper goes one step further, as it presents a method to extract TRIZ contradictions from patent texts based on Prompt Engineering using a generative Large Language Model (LLM), namely OpenAI's GPT-4. Contradiction detection, sentence extraction, contradiction summarization, parameter extraction and assignment to the 39 abstract TRIZ engineering parameters are all performed in a single prompt using the LangChain framework. Our results show that "off-the-shelf" GPT-4 is a serious alternative to existing approaches.
Improving the accessibility of psychotherapy with the aid of Large Language Models (LLMs) is garnering a significant attention in recent years. Recognizing cognitive distortions from the interviewee's utterances can be an essential part of psychotherapy, especially for cognitive behavioral therapy. In this paper, we propose ERD, which improves LLM-based cognitive distortion classification performance with the aid of additional modules of (1) extracting the parts related to cognitive distortion, and (2) debating the reasoning steps by multiple agents. Our experimental results on a public dataset show that ERD improves the multi-class F1 score as well as binary specificity score. Regarding the latter score, it turns out that our method is effective in debiasing the baseline method which has high false positive rate, especially when the summary of multi-agent debate is provided to LLMs.
Despite the promise of RLHF in aligning LLMs with human preferences, it often leads to superficial alignment, prioritizing stylistic changes over improving downstream performance of LLMs. Underspecified preferences could obscure directions to align the models. Lacking exploration restricts identification of desirable outputs to improve the models. To overcome these challenges, we propose a novel framework: Reinforcement Learning from Reflective Feedback (RLRF), which leverages fine-grained feedback based on detailed criteria to improve the core capabilities of LLMs. RLRF employs a self-reflection mechanism to systematically explore and refine LLM responses, then fine-tuning the models via a RL algorithm along with promising responses. Our experiments across Just-Eval, Factuality, and Mathematical Reasoning demonstrate the efficacy and transformative potential of RLRF beyond superficial surface-level adjustment.
Despite the remarkable performance of video-based large language models (LLMs), their adversarial threat remains unexplored. To fill this gap, we propose the first adversarial attack tailored for video-based LLMs by crafting flow-based multi-modal adversarial perturbations on a small fraction of frames within a video, dubbed FMM-Attack. Extensive experiments show that our attack can effectively induce video-based LLMs to generate incorrect answers when videos are added with imperceptible adversarial perturbations. Intriguingly, our FMM-Attack can also induce garbling in the model output, prompting video-based LLMs to hallucinate. Overall, our observations inspire a further understanding of multi-modal robustness and safety-related feature alignment across different modalities, which is of great importance for various large multi-modal models. Our code is available at https://github.com/THU-Kingmin/FMM-Attack.
In children's collaborative learning, effective peer conversations can significantly enhance the quality of children's collaborative interactions. The integration of Large Language Model (LLM) agents into this setting explores their novel role as peers, assessing impacts as team moderators and participants. We invited two groups of participants to engage in a collaborative learning workshop, where they discussed and proposed conceptual solutions to a design problem. The peer conversation transcripts were analyzed using thematic analysis. We discovered that peer agents, while managing discussions effectively as team moderators, sometimes have their instructions disregarded. As participants, they foster children's creative thinking but may not consistently provide timely feedback. These findings highlight potential design improvements and considerations for peer agents in both roles.
The socially-aware navigation system has evolved to adeptly avoid various obstacles while performing multiple tasks, such as point-to-point navigation, human-following, and -guiding. However, a prominent gap persists: in Human-Robot Interaction (HRI), the procedure of communicating commands to robots demands intricate mathematical formulations. Furthermore, the transition between tasks does not quite possess the intuitive control and user-centric interactivity that one would desire. In this work, we propose an LLM-driven interactive multimodal multitask robot navigation framework, termed LIM2N, to solve the above new challenge in the navigation field. We achieve this by first introducing a multimodal interaction framework where language and hand-drawn inputs can serve as navigation constraints and control objectives. Next, a reinforcement learning agent is built to handle multiple tasks with the received information. Crucially, LIM2N creates smooth cooperation among the reasoning of multimodal input, multitask planning, and adaptation and processing of the intelligent sensing modules in the complicated system. Extensive experiments are conducted in both simulation and the real world demonstrating that LIM2N has superior user needs understanding, alongside an enhanced interactive experience.
Instruction tuning is crucial for enabling Language Learning Models (LLMs) in responding to human instructions. The quality of instruction pairs used for tuning greatly affects the performance of LLMs. However, the manual creation of high-quality instruction datasets is costly, leading to the adoption of automatic generation of instruction pairs by LLMs as a popular alternative. To ensure the high quality of LLM-generated instruction datasets, several approaches have been proposed. Nevertheless, existing methods either compromise dataset integrity by filtering a large proportion of samples, or are unsuitable for industrial applications. In this paper, instead of discarding low-quality samples, we propose CoachLM, a novel approach to enhance the quality of instruction datasets through automatic revisions on samples in the dataset. CoachLM is trained from the samples revised by human experts and significantly increases the proportion of high-quality samples in the dataset from 17.7% to 78.9%. The effectiveness of CoachLM is further assessed on various real-world instruction test sets. The results show that CoachLM improves the instruction-following capabilities of the instruction-tuned LLM by an average of 29.9%, which even surpasses larger LLMs with nearly twice the number of parameters. Furthermore, CoachLM is successfully deployed in a data management system for LLMs at Huawei, resulting in an efficiency improvement of up to 20% in the cleaning of 40k real-world instruction pairs. We release various assets of CoachLM, including the training data, code and test set (https://github.com/lunyiliu/CoachLM).
Ensuring alignment, which refers to making models behave in accordance with human intentions [1,2], has become a critical task before deploying large language models (LLMs) in real-world applications. For instance, OpenAI devoted six months to iteratively aligning GPT-4 before its release [3]. However, a major challenge faced by practitioners is the lack of clear guidance on evaluating whether LLM outputs align with social norms, values, and regulations. This obstacle hinders systematic iteration and deployment of LLMs. To address this issue, this paper presents a comprehensive survey of key dimensions that are crucial to consider when assessing LLM trustworthiness. The survey covers seven major categories of LLM trustworthiness: reliability, safety, fairness, resistance to misuse, explainability and reasoning, adherence to social norms, and robustness. Each major category is further divided into several sub-categories, resulting in a total of 29 sub-categories. Additionally, a subset of 8 sub-categories is selected for further investigation, where corresponding measurement studies are designed and conducted on several widely-used LLMs. The measurement results indicate that, in general, more aligned models tend to perform better in terms of overall trustworthiness. However, the effectiveness of alignment varies across the different trustworthiness categories considered. This highlights the importance of conducting more fine-grained analyses, testing, and making continuous improvements on LLM alignment. By shedding light on these key dimensions of LLM trustworthiness, this paper aims to provide valuable insights and guidance to practitioners in the field. Understanding and addressing these concerns will be crucial in achieving reliable and ethically sound deployment of LLMs in various applications.
This study targets a critical aspect of multi-modal LLMs' (LLMs&VLMs) inference: explicit controllable text generation. Multi-modal LLMs empower multi-modality understanding with the capability of semantic generation yet bring less explainability and heavier reliance on prompt contents due to their autoregressive generative nature. While manipulating prompt formats could improve outputs, designing specific and precise prompts per task can be challenging and ineffective. To tackle this issue, we introduce a novel inference method, Prompt Highlighter, which enables users to highlight specific prompt spans to interactively control the focus during generation. Motivated by the classifier-free diffusion guidance, we form regular and unconditional context pairs based on highlighted tokens, demonstrating that the autoregressive generation in models can be guided in a classifier-free way. Notably, we find that, during inference, guiding the models with highlighted tokens through the attention weights leads to more desired outputs. Our approach is compatible with current LLMs and VLMs, achieving impressive customized generation results without training. Experiments confirm its effectiveness in focusing on input contexts and generating reliable content. Without tuning on LLaVA-v1.5, our method secured 70.7 in the MMBench test and 1552.5 in MME-perception. The code is available at: https://github.com/dvlab-research/Prompt-Highlighter/
Developing novel research questions (RQs) often requires extensive literature reviews, especially in interdisciplinary fields. To support RQ development through human-AI co-creation, we leveraged Large Language Models (LLMs) to build an LLM-based agent system named CoQuest. We conducted an experiment with 20 HCI researchers to examine the impact of two interaction designs: breadth-first and depth-first RQ generation. The findings revealed that participants perceived the breadth-first approach as more creative and trustworthy upon task completion. Conversely, during the task, participants considered the depth-first generated RQs as more creative. Additionally, we discovered that AI processing delays allowed users to reflect on multiple RQs simultaneously, leading to a higher quantity of generated RQs and an enhanced sense of control. Our work makes both theoretical and practical contributions by proposing and evaluating a mental model for human-AI co-creation of RQs. We also address potential ethical issues, such as biases and over-reliance on AI, advocating for using the system to improve human research creativity rather than automating scientific inquiry.
Large Language Models (LLMs), such as ChatGPT and Bard, have revolutionized natural language understanding and generation. They possess deep language comprehension, human-like text generation capabilities, contextual awareness, and robust problem-solving skills, making them invaluable in various domains (e.g., search engines, customer support, translation). In the meantime, LLMs have also gained traction in the security community, revealing security vulnerabilities and showcasing their potential in security-related tasks. This paper explores the intersection of LLMs with security and privacy. Specifically, we investigate how LLMs positively impact security and privacy, potential risks and threats associated with their use, and inherent vulnerabilities within LLMs. Through a comprehensive literature review, the paper categorizes the papers into "The Good" (beneficial LLM applications), "The Bad" (offensive applications), and "The Ugly" (vulnerabilities of LLMs and their defenses). We have some interesting findings. For example, LLMs have proven to enhance code security (code vulnerability detection) and data privacy (data confidentiality protection), outperforming traditional methods. However, they can also be harnessed for various attacks (particularly user-level attacks) due to their human-like reasoning abilities. We have identified areas that require further research efforts. For example, Research on model and parameter extraction attacks is limited and often theoretical, hindered by LLM parameter scale and confidentiality. Safe instruction tuning, a recent development, requires more exploration. We hope that our work can shed light on the LLMs' potential to both bolster and jeopardize cybersecurity.
The abilities of large language models (LLMs) have recently progressed to unprecedented levels, paving the way to novel applications in a wide variety of areas. In computer vision, LLMs can be used to prime vision-language tasks such image captioning and visual question answering when coupled with pre-trained vision backbones. While different approaches have been explored to interface LLMs with ``perceptual backbones'' that process, e.g., visual or audio data, they are often explored for different tasks, different datasets, and using different perceptual backbones and language models, hindering direct comparison of the interfacing mechanisms. To remedy this lack of comparability between methods, we present an extensive experimental evaluation of different interfacing mechanisms, across multiple tasks (including image, video, and audio captioning as well as visual question answering), datasets and backbones, paying special attention to low-data settings. We find improved performance using existing mechanisms over state-of-the-art results, and identify a new interfacing mechanism that yields (near) optimal results across different tasks, while obtaining a 4x reduction in training time.
Automatic Chinese classical poetry generation has attracted much research interest, but achieving effective control over format and content simultaneously remains challenging. Traditional systems usually accept keywords as user inputs, resulting in limited control over content. Large language models (LLMs) improve content control by allowing unrestricted user instructions, but the token-by-token generation process frequently makes format errors. Motivated by this, we propose CharPoet, a Chinese classical poetry generation system based on token-free LLM, which provides effective control over both format and content. Our token-free architecture generates in a character-by-character manner, enabling precise control over the number of characters. Pruned from existing token-based LLMs, CharPoet inherits their pretrained capabilities and can generate poetry following instructions like "Write me a poem for my mother's birthday." CharPoet achieves format accuracy above 0.96, outperforming Jiuge-GPT-2 (0.91) and GPT-4 (0.38). In terms of content quality, CharPoet surpasses traditional systems including Jiuge, and is comparable to other LLMs. Our system is open source and available at https://modelscope.cn/models/CharPoet/CharPoet. A video demonstration of CharPoet is available at https://youtu.be/voZ25qEp3Dc.
Large language models (LLMs) have reached human-like proficiency in generating diverse textual content, underscoring the necessity for effective fake text detection to avoid potential risks such as fake news in social media. Previous research has mostly tested single models on in-distribution datasets, limiting our understanding of how these models perform on different types of data for LLM-generated text detection task. We researched this by testing five specialized transformer-based models on both in-distribution and out-of-distribution datasets to better assess their performance and generalizability. Our results revealed that single transformer-based classifiers achieved decent performance on in-distribution dataset but limited generalization ability on out-of-distribution dataset. To improve it, we combined the individual classifiers models using adaptive ensemble algorithms, which improved the average accuracy significantly from 91.8% to 99.2% on an in-distribution test set and from 62.9% to 72.5% on an out-of-distribution test set. The results indicate the effectiveness, good generalization ability, and great potential of adaptive ensemble algorithms in LLM-generated text detection.
We develop Polaris, the first safety-focused LLM constellation for real-time patient-AI healthcare conversations. Unlike prior LLM works in healthcare focusing on tasks like question answering, our work specifically focuses on long multi-turn voice conversations. Our one-trillion parameter constellation system is composed of several multibillion parameter LLMs as co-operative agents: a stateful primary agent that focuses on driving an engaging conversation and several specialist support agents focused on healthcare tasks performed by nurses to increase safety and reduce hallucinations. We develop a sophisticated training protocol for iterative co-training of the agents that optimize for diverse objectives. We train our models on proprietary data, clinical care plans, healthcare regulatory documents, medical manuals, and other medical reasoning documents. We align our models to speak like medical professionals, using organic healthcare conversations and simulated ones between patient actors and experienced nurses. This allows our system to express unique capabilities such as rapport building, trust building, empathy and bedside manner. Finally, we present the first comprehensive clinician evaluation of an LLM system for healthcare. We recruited over 1100 U.S. licensed nurses and over 130 U.S. licensed physicians to perform end-to-end conversational evaluations of our system by posing as patients and rating the system on several measures. We demonstrate Polaris performs on par with human nurses on aggregate across dimensions such as medical safety, clinical readiness, conversational quality, and bedside manner. Additionally, we conduct a challenging task-based evaluation of the individual specialist support agents, where we demonstrate our LLM agents significantly outperform a much larger general-purpose LLM (GPT-4) as well as from its own medium-size class (LLaMA-2 70B).
The rapid integration of Large Language Models (LLMs) across diverse sectors has marked a transformative era, showcasing remarkable capabilities in text generation and problem-solving tasks. However, this technological advancement is accompanied by significant risks and vulnerabilities. Despite ongoing security enhancements, attackers persistently exploit these weaknesses, casting doubts on the overall trustworthiness of LLMs. Compounding the issue, organisations are deploying LLM-integrated systems without understanding the severity of potential consequences. Existing studies by OWASP and MITRE offer a general overview of threats and vulnerabilities but lack a method for directly and succinctly analysing the risks for security practitioners, developers, and key decision-makers who are working with this novel technology. To address this gap, we propose a risk assessment process using tools like the OWASP risk rating methodology which is used for traditional systems. We conduct scenario analysis to identify potential threat agents and map the dependent system components against vulnerability factors. Through this analysis, we assess the likelihood of a cyberattack. Subsequently, we conduct a thorough impact analysis to derive a comprehensive threat matrix. We also map threats against three key stakeholder groups: developers engaged in model fine-tuning, application developers utilizing third-party APIs, and end users. The proposed threat matrix provides a holistic evaluation of LLM-related risks, enabling stakeholders to make informed decisions for effective mitigation strategies. Our outlined process serves as an actionable and comprehensive tool for security practitioners, offering insights for resource management and enhancing the overall system security.
In human-centered design, developing a comprehensive and in-depth understanding of user experiences, i.e., empathic understanding, is paramount for designing products that truly meet human needs. Nevertheless, accurately comprehending the real underlying mental states of a large human population remains a significant challenge today. This difficulty mainly arises from the trade-off between depth and scale of user experience research: gaining in-depth insights from a small group of users does not easily scale to a larger population, and vice versa. This paper investigates the use of Large Language Models (LLMs) for performing mental inference tasks, specifically inferring users' underlying goals and fundamental psychological needs (FPNs). Baseline and benchmark datasets were collected from human users and designers to develop an empathic accuracy metric for measuring the mental inference performance of LLMs. The empathic accuracy of inferring goals and FPNs of different LLMs with varied zero-shot prompt engineering techniques are experimented against that of human designers. Experimental results suggest that LLMs can infer and understand the underlying goals and FPNs of users with performance comparable to that of human designers, suggesting a promising avenue for enhancing the scalability of empathic design approaches through the integration of advanced artificial intelligence technologies. This work has the potential to significantly augment the toolkit available to designers during human-centered design, enabling the development of both large-scale and in-depth understanding of users' experiences.
Large Language Models (LLMs) have recently made significant advances in code generation through the 'Chain-of-Thought' prompting technique. This technique empowers the model to autonomously devise "solution plans" to tackle intricate programming challenges, thereby improving its performance in code generation. Nevertheless, smaller models have been struggling to keep up with LLMs in deducing these plans, adversely affecting their code generation capabilities. Given the considerable size and associated deployment costs, along with concerns about data security, many teams opt for deploying smaller models for code generation. Consequently, there arises a compelling need for transferring LLMs' code generation reasoning abilities to the smaller models. In this paper, we propose the CodePLAN framework, which aims to transfer LLMs' reasoning capabilities to smaller models through distillation. We adopt a multi-task learning approach, jointly undertaking code generation and solution plan generation tasks, to enhance the code generation capabilities of the smaller model. To ensure the superior quality of the solution plans, we advocate for the utilization of backward reasoning and plan sampling strategies. Our experiments show that in comparison to the conventional fine-tuning approach, our approach improves the smaller model's code generation performance (measured in pass@1 metric) by over 130% on the challenging APPS benchmark.
Electronic health records (EHRs) contain valuable patient data for health-related prediction tasks, such as disease prediction. Traditional approaches rely on supervised learning methods that require large labeled datasets, which can be expensive and challenging to obtain. In this study, we investigate the feasibility of applying Large Language Models (LLMs) to convert structured patient visit data (e.g., diagnoses, labs, prescriptions) into natural language narratives. We evaluate the zero-shot and few-shot performance of LLMs using various EHR-prediction-oriented prompting strategies. Furthermore, we propose a novel approach that utilizes LLM agents with different roles: a predictor agent that makes predictions and generates reasoning processes and a critic agent that analyzes incorrect predictions and provides guidance for improving the reasoning of the predictor agent. Our results demonstrate that with the proposed approach, LLMs can achieve decent few-shot performance compared to traditional supervised learning methods in EHR-based disease predictions, suggesting its potential for health-oriented applications.
Watermarking approaches are proposed to identify if text being circulated is human or large language model (LLM) generated. The state-of-the-art watermarking strategy of Kirchenbauer et al. (2023a) biases the LLM to generate specific (``green'') tokens. However, determining the robustness of this watermarking method is an open problem. Existing attack methods fail to evade detection for longer text segments. We overcome this limitation, and propose {\em Self Color Testing-based Substitution (SCTS)}, the first ``color-aware'' attack. SCTS obtains color information by strategically prompting the watermarked LLM and comparing output tokens frequencies. It uses this information to determine token colors, and substitutes green tokens with non-green ones. In our experiments, SCTS successfully evades watermark detection using fewer number of edits than related work. Additionally, we show both theoretically and empirically that SCTS can remove the watermark for arbitrarily long watermarked text.
In this paper, we introduce Semantic Layering in Room Segmentation via LLMs (SeLRoS), an advanced method for semantic room segmentation by integrating Large Language Models (LLMs) with traditional 2D map-based segmentation. Unlike previous approaches that solely focus on the geometric segmentation of indoor environments, our work enriches segmented maps with semantic data, including object identification and spatial relationships, to enhance robotic navigation. By leveraging LLMs, we provide a novel framework that interprets and organizes complex information about each segmented area, thereby improving the accuracy and contextual relevance of room segmentation. Furthermore, SeLRoS overcomes the limitations of existing algorithms by using a semantic evaluation method to accurately distinguish true room divisions from those erroneously generated by furniture and segmentation inaccuracies. The effectiveness of SeLRoS is verified through its application across 30 different 3D environments. Source code and experiment videos for this work are available at: https://sites.google.com/view/selros.
Large language models (LLM) have recently emerged as a powerful tool for a variety of natural language processing tasks, bringing a new surge of combining LLM with recommendation systems, termed as LLM-based RS. Current approaches generally fall into two main paradigms, the ID direct usage paradigm and the ID translation paradigm, noting their core weakness stems from lacking recommendation knowledge and uniqueness. To address this limitation, we propose a new paradigm, ID representation, which incorporates pre-trained ID embeddings into LLMs in a complementary manner. In this work, we present RA-Rec, an efficient ID representation alignment framework for LLM-based recommendation, which is compatible with multiple ID-based methods and LLM architectures. Specifically, we treat ID embeddings as soft prompts and design an innovative alignment module and an efficient tuning method with tailored data construction for alignment. Extensive experiments demonstrate RA-Rec substantially outperforms current state-of-the-art methods, achieving up to 3.0% absolute HitRate@100 improvements while utilizing less than 10x training data.