Large language models (LLMs) have achieved remarkable performance on various NLP tasks and are augmented by tools for broader applications. Yet, how to evaluate and analyze the tool utilization capability of LLMs is still under-explored. In contrast to previous works that evaluate models holistically, we comprehensively decompose the tool utilization into multiple sub-processes, including instruction following, planning, reasoning, retrieval, understanding, and review. Based on that, we further introduce T-Eval to evaluate the tool-utilization capability step by step. T-Eval disentangles the tool utilization evaluation into several sub-domains along model capabilities, facilitating the inner understanding of both holistic and isolated competency of LLMs. We conduct extensive experiments on T-Eval and in-depth analysis of various LLMs. T-Eval not only exhibits consistency with the outcome-oriented evaluation but also provides a more fine-grained analysis of the capabilities of LLMs, providing a new perspective in LLM evaluation on tool-utilization ability. The benchmark will be available.
论文检索
输入标题、作者或关键词，从 1,915 篇学术成果中精准定位
Fine-Grained Image-Text Alignment in Medical Imaging Enables Explainable Cyclic Image-Report Generation
PDF ↗Fine-grained vision-language models (VLM) have been widely used for inter-modality local alignment between the predefined fixed patches and textual words. However, in medical analysis, lesions exhibit varying sizes and positions, and using fixed patches may cause incomplete representations of lesions. Moreover, these methods provide explainability by using heatmaps to show the general image areas potentially associated with texts rather than specific regions, making their explanations not explicit and specific enough. To address these issues, we propose a novel Adaptive patch-word Matching (AdaMatch) model to correlate chest X-ray (CXR) image regions with words in medical reports and apply it to CXR-report generation to provide explainability for the generation process. AdaMatch exploits the fine-grained relation between adaptive patches and words to provide explanations of specific image regions with corresponding words. To capture the abnormal regions of varying sizes and positions, we introduce an Adaptive Patch extraction (AdaPatch) module to acquire adaptive patches for these regions adaptively. Aiming to provide explicit explainability for the CXR-report generation task, we propose an AdaMatch-based bidirectional LLM for Cyclic CXR-report generation (AdaMatch-Cyclic). It employs AdaMatch to obtain the keywords for CXR images and ‘keypatches’ for medical reports as hints to guide CXR-report generation. Extensive experiments on two publicly available CXR datasets validate the effectiveness of our method and its superior performance over existing methods. Source code will be released.
Video storytelling is engaging multimedia content that utilizes video and its accompanying narration to share a story and attract the audience, where a key challenge is creating narrations for recorded visual scenes. Previous studies on dense video captioning and video story generation have made some progress. However, in practical applications, we typically require synchronized narrations for ongoing visual scenes. In this work, we introduce a new task of Synchronized Video Storytelling, which aims to generate synchronous and informative narrations for videos. These narrations, associated with each video clip, should relate to the visual content, integrate relevant knowledge, and have an appropriate word count corresponding to the clip’s duration. Specifically, a structured storyline is beneficial to guide the generation process, ensuring coherence and integrity. To support the exploration of this task, we introduce a new benchmark dataset E-SyncVidStory with rich annotations. Since existing Multimodal LLMs are not effective in addressing this task in one-shot or few-shot settings, we propose a framework named VideoNarrator that can generate a storyline for input videos and simultaneously generate narrations with the guidance of the generated or predefined storyline. We further introduce a set of evaluation metrics to thoroughly assess the generation. Both automatic and human evaluations validate the effectiveness of our approach. Our dataset, codes, and evaluations will be released.
Event relations are crucial for narrative understanding and reasoning. Governed by nuanced logic, event relation extraction (ERE) is a challenging task that demands thorough semantic understanding and rigorous logical reasoning. In this paper, we conduct an in-depth investigation to systematically explore the capability of LLMs in understanding and applying event relation logic. More in detail, we first investigate the deficiencies of LLMs in logical reasoning across different tasks. Our study reveals that LLMs are not logically consistent reasoners, which results in their suboptimal performance on tasks that need rigorous reasoning. To address this, we explore three different approaches to endow LLMs with event relation logic, and thus enable them to generate more coherent answers across various scenarios. Based on our approach, we also contribute a synthesized dataset (LLM-ERL) involving high-order reasoning for evaluation and fine-tuning. Extensive quantitative and qualitative analyses on different tasks also validate the effectiveness of our approach and provide insights for solving practical tasks with LLMs in future work. Codes are available at https://github.com/chenmeiqii/Teach-LLM-LR.
In this paper, we uncover a positional bias in the evaluation paradigm of adopting large language models (LLMs), e.g., GPT-4, as a referee to score and compare the quality of responses generated by candidate models. We find that the quality ranking of candidate responses can be easily hacked by simply altering their order of appearance in the context. This manipulation allows us to skew the evaluation result, making one model appear considerably superior to the other, e.g., Vicuna-13B could beat ChatGPT on 66 over 80 tested queries with ChatGPT as an evaluator. We propose a simple yet effective calibration framework to address our discovered positional bias.To evaluate the effectiveness of our framework, we manually annotate the “win/tie/lose” outcomes of responses from ChatGPT and Vicuna-13B in the Vicuna Benchmark’s question prompt. Extensive experiments demonstrate that our approach successfully alleviates evaluation bias, resulting in closer alignment with human judgments.
In this paper, we present an innovative process-oriented math process reward model called Math-shepherd, which assigns a reward score to each step of math problem solutions. The training of Math-shepherd is achieved using automatically constructed process-wise supervision data, breaking the bottleneck of heavy reliance on manual annotation in existing work. We explore the effectiveness of Math-shepherd in two scenarios: 1) \textit{Verification}: Math-shepherd is utilized for reranking multiple outputs generated by Large Language Models (LLMs); 2) \textit{Reinforcement Learning (RL)}: Math-shepherd is employed to reinforce LLMs.With Math-shepherd, a series of open-source LLMs demonstrates exceptional performance. For instance, process RL with Math-shepherd significantly enhances Mistral-7B (77.9%\to84.1% on GSM8K and 28.6%\to33.0% on MATH).The accuracy can be further improved to 89.1% and 43.5% on two benchmarks with verification of Math-shepherd.We believe that automatic process supervision holds significant potential for the future evolution of LLMs.
Whose Preferences? Differences in Fairness Preferences and Their Impact on the Fairness of AI Utilizing Human Feedback
PDF ↗There is a growing body of work on learning from human feedback to align various aspects of machine learning systems with human values and preferences. We consider the setting of fairness in content moderation, in which human feedback is used to determine how two comments — referencing different sensitive attribute groups — should be treated in comparison to one another. With a novel dataset collected from Prolific and MTurk, we find significant gaps in fairness preferences depending on the race, age, political stance, educational level, and LGBTQ+ identity of annotators. We also demonstrate that demographics mentioned in text have a strong influence on how users perceive individual fairness in moderation. Further, we find that differences also exist in downstream classifiers trained to predict human preferences. Finally, we observe that an ensemble, giving equal weight to classifiers trained on annotations from different demographics, performs better for different demographic intersections; compared to a single classifier that gives equal weight to each annotation.
Deductive reasoning plays a pivotal role in the formulation of sound and cohesive arguments. It allows individuals to draw conclusions that logically follow, given the truth value of the information provided. Recent progress in the domain of large language models (LLMs) has showcased their capability in executing deductive reasoning tasks. Nonetheless, a significant portion of research primarily assesses the accuracy of LLMs in solving such tasks, often overlooking a deeper analysis of their reasoning behavior. In this study, we draw upon principles from cognitive psychology to examine inferential strategies employed by LLMs, through a detailed evaluation of their responses to propositional logic problems. Our findings indicate that LLMs display reasoning patterns akin to those observed in humans, including strategies like \textit{supposition following} or \textit{chain construction}. Moreover, our research demonstrates that the architecture and scale of the model significantly affect its preferred method of reasoning, with more advanced models tending to adopt strategies more frequently than less sophisticated ones. Importantly, we assert that a model’s accuracy, that is the correctness of its final conclusion, does not necessarily reflect the validity of its reasoning process. This distinction underscores the necessity for more nuanced evaluation procedures in the field.
Large language models (LLMs) garner significant attention for their unprecedented performance, leading to an increasing number of researches evaluating LLMs. However, these evaluation benchmarks are limited to assessing the instruction-following capabilities, overlooking the fundamental abilities that emerge during the pre-training stage. Previous subjective evaluation methods mainly reply on scoring by API models. However, in the absence of references, large models have shown limited ability to discern subtle differences. To bridge the gap, we propose F-Eval, a bilingual evaluation benchmark to evaluate the fundamental abilities, including expression, commonsense and logic. The tasks in F-Eval include multi-choice objective tasks, open-ended objective tasks, reference-based subjective tasks and reference-free subjective tasks. For reference-free subjective tasks, we devise new evaluation methods, serving as alternatives to scoring by API models. We conduct evaluations on 13 advanced LLMs. Results show that our evaluation methods show higher correlation coefficients and larger distinction than other evaluators. Additionally, we discuss the influence of different model sizes, dimensions, and normalization methods. We anticipate that F-Eval will facilitate the study of LLMs’ fundamental abilities.
Despite the many advances of Large Language Models (LLMs) and their unprecedented rapid evolution, their impact and integration into every facet of our daily lives is limited due to various reasons. One critical factor hindering their widespread adoption is the occurrence of hallucinations, where LLMs invent answers that sound realistic, yet drift away from factual truth. In this paper, we present a novel method for detecting hallucinations in large language models, which tackles a critical issue in the adoption of these models in various real-world scenarios. Through extensive evaluations across multiple datasets and LLMs, including Llama-2, we study the hallucination levels of various recent LLMs and demonstrate the effectiveness of our method to automatically detect them. Notably, we observe up to 87% hallucinations for Llama-2 in a specific experiment, where our method achieves a Balanced Accuracy of 81%, all without relying on external knowledge.
Graphical User Interface (GUI) agents are designed to automate complex tasks on digital devices, such as smartphones and desktops. Most existing GUI agents interact with the environment through extracted structured data, which can be notably lengthy (e.g., HTML) and occasionally inaccessible (e.g., on desktops). To alleviate this issue, we propose a novel visual GUI agent – SeeClick, which only relies on screenshots for task automation. In our preliminary study, we have discovered a key challenge in developing visual GUI agents: GUI grounding – the capacity to accurately locate screen elements based on instructions. To tackle this challenge, we propose to enhance SeeClick with GUI grounding pre-training and devise a method to automate the curation of GUI grounding data. Along with the efforts above, we have also created ScreenSpot, the first realistic GUI grounding benchmark that encompasses mobile, desktop, and web environments. After pre-training, SeeClick demonstrates significant improvement in ScreenSpot over various baselines. Moreover, comprehensive evaluations on three widely used benchmarks consistently support our finding that advancements in GUI grounding directly correlate with enhanced performance in downstream GUI agent tasks. The model, data and code will be open-sourced.
Spoken language glossification (SLG) aims to translate the spoken language text into the sign language gloss, i.e., a written record of sign language. In this work, we present a framework named Semi-Supervised Spoken Language Glossification (S^3LG) for SLG. To tackle the bottleneck of limited parallel data in SLG, our S^3LG incorporates large-scale monolingual spoken language text into SLG training. The proposed framework follows the self-training structure that iteratively annotates and learns from pseudo labels. Considering the lexical similarity and syntactic difference between sign language and spoken language, our S^3LG adopts both the rule-based heuristic and model-based approach for auto-annotation. During training, we randomly mix these complementary synthetic datasets and mark their differences with a special token. As the synthetic data may be less quality, the S^3LG further leverages consistency regularization to reduce the negative impact of noise in the synthetic data. Extensive experiments are conducted on public benchmarks to demonstrate the effectiveness of the S^3LG. Our code is available at https://github.com/yaohj11/S3LG.
Recent advances in retrieval-augmented models for image captioning highlight the benefit of retrieving related captions for efficient, lightweight models with strong domain-transfer capabilities. While these models demonstrate the success of retrieval augmentation, retrieval models are still far from perfect in practice: the retrieved information can sometimes mislead the model, resulting in incorrect generation and worse performance. In this paper, we analyze the robustness of a retrieval-augmented captioning model SmallCap. Our analysis shows that the model is sensitive to tokens that appear in the majority of the retrieved captions, and the input attribution shows that those tokens are likely copied into the generated output. Given these findings, we propose to train the model by sampling retrieved captions from more diverse sets. This decreases the chance that the model learns to copy majority tokens, and improves both in-domain and cross-domain performance.
Event Argument Extraction (EAE) aims to extract arguments for specified events from a text. Previous research has mainly focused on addressing long-distance dependencies of arguments, modeling co-occurrence relationships between roles and events, but overlooking potential inductive biases: (i) semantic differences among arguments of the same type and (ii) large margin separation between arguments of the different types. Inspired by prototype networks, we introduce a new model named HMPEAE, which takes the two inductive biases above as targets to locate prototypes and guide the model to learn argument representations based on these prototypes.Specifically, we set multiple prototypes to represent each role to capture intra-class differences. Simultaneously, we use hypersphere as the output space for prototypes, defining large margin separation between prototypes to encourage the model to learn significant differences between different types of arguments effectively.We solve the “argument-prototype” assignment as an optimal transport problem to optimize the argument representation and minimize the absolute distance between arguments and prototypes to achieve compactness within sub-clusters. Experimental results on the RAMS and WikiEvents datasets show that HMPEAE achieves state-of-the-art performances.
Relation extraction (RE) is a fundamental task in natural language processing, aiming to identify relations between target entities in text. While many RE methods are designed for a single sentence or document, cross-document RE has emerged to address relations across multiple long documents. Given the nature of long documents in cross-document RE, extracting document embeddings is challenging due to the length constraints of pre-trained language models. Therefore, we propose REward-based Input Construction (REIC), the first learning-based sentence selector for cross-document RE. REIC extracts sentences based on relational evidence, enabling the RE module to effectively infer relations. Since supervision of evidence sentences is generally unavailable, we train REIC using reinforcement learning with RE prediction scores as rewards. Experimental results demonstrate the superiority of our method over heuristic methods for different RE structures and backbones in cross-document RE. Our code is publicly available at https://github.com/aailabkaist/REIC.
Multi-aspect controllable text generation aims to control the generated texts in attributes from multiple aspects (e.g., “positive” from sentiment and “sport” from topic). Existing works neglect attribute correlations formed by the intertwining of different attributes. Particularly, the stereotype formed by imbalanced attribute correlations significantly affects multi-aspect control. In this paper, we propose MAGIC, a new multi-aspect controllable text generation method with disentangled counterfactual augmentation. We alleviate the issue of imbalanced attribute correlations during training using counterfactual feature vectors in the attribute latent space by disentanglement. During inference, we enhance attribute correlations by target-guided counterfactual augmentation to further improve multi-aspect control. Experiments show that MAGIC outperforms state-of-the-art baselines in both imbalanced and balanced attribute correlation scenarios.
Focus on Your Question! Interpreting and Mitigating Toxic CoT Problems in Commonsense Reasoning
PDF ↗Large language models exhibit high-level commonsense reasoning abilities, especially with enhancement methods like Chain-of-Thought (CoT). However, we find these CoT-like methods lead to a considerable number of originally correct answers turning wrong, which we define as the Toxic CoT problem. To interpret and mitigate this problem, we first utilize attribution tracing and causal tracing methods to probe the internal working mechanism of the LLM during CoT reasoning. Through comparisons, we prove that the model exhibits information loss from the question over the shallow attention layers when generating rationales or answers. Based on the probing findings, we design a novel method called RIDERS (Residual decodIng and sERial-position Swap), which compensates for the information deficit in the model from both decoding and serial-position perspectives. Through extensive experiments on multiple commonsense reasoning benchmarks, we validate that this method not only significantly eliminates Toxic CoT problems (decreased by \textbf{23.6}%), but also effectively improves the model’s overall commonsense reasoning performance (increased by \textbf{5.5}%).
MM-SAP: A Comprehensive Benchmark for Assessing Self-Awareness of Multimodal Large Language Models in Perception
PDF ↗Recent advancements in Multimodal Large Language Models (MLLMs) have demonstrated exceptional capabilities in visual perception and understanding. However, these models also suffer from hallucinations, which limit their reliability as AI systems. We believe that these hallucinations are partially due to the models’ struggle with understanding what they can and cannot perceive from images, a capability we refer to as self-awareness in perception. Despite its importance, this aspect of MLLMs has been overlooked in prior studies. In this paper, we aim to define and evaluate the self-awareness of MLLMs in perception. To do this, we first introduce the knowledge quadrant in perception, which helps define what MLLMs know and do not know about images. Using this framework, we propose a novel benchmark, the Self-Awareness in Perception for MLLMs (MM-SAP), specifically designed to assess this capability. We apply MM-SAP to a variety of popular MLLMs, offering a comprehensive analysis of their self-awareness and providing detailed insights. The experiment results reveal that current MLLMs possess limited self-awareness capabilities, pointing to a crucial area for future advancement in the development of trustworthy MLLMs. Code and data are available at https://github.com/YHWmz/MM-SAP.
When adapting Large Language Models for Recommendation (LLMRec), it is crucial to integrate collaborative information. Existing methods achieve this by learning collaborative embeddings in LLMs’ latent space from scratch or by mapping from external models. However, they fail to represent the information in a text-like format, which may not align optimally with LLMs. To bridge this gap, we introduce BinLLM, a novel LLMRec method that seamlessly integrates collaborative information through text-like encoding. BinLLM converts collaborative embeddings from external models into binary sequences — a specific text format that LLMs can understand and operate on directly, facilitating the direct usage of collaborative information in text-like format by LLMs. Additionally, BinLLM provides options to compress the binary sequence using dot-decimal notation to avoid excessively long lengths. Extensive experiments validate that BinLLM introduces collaborative information in a manner better aligned with LLMs, resulting in enhanced performance. We release our code at https://github.com/zyang1580/BinLLM.
Text watermarking has emerged as a pivotal technique for identifying machine-generated text. However, existing methods often rely on arbitrary vocabulary partitioning during decoding to embed watermarks, which compromises the availability of suitable tokens and significantly degrades the quality of responses. This study assesses the impact of watermarking on different capabilities of large language models (LLMs) from a cognitive science lens. Our finding highlights a significant disparity; knowledge recall and logical reasoning are more adversely affected than language generation. These results suggest a more profound effect of watermarking on LLMs than previously understood. To address these challenges, we introduce Watermarking with Mutual Exclusion (WatME), a novel approach leveraging linguistic prior knowledge of inherent lexical redundancy in LLM vocabularies to seamlessly integrate watermarks. Specifically, WatME dynamically optimizes token usage during the decoding process by applying a mutually exclusive rule to the identified lexical redundancies. This strategy effectively prevents the unavailability of appropriate tokens and preserves the expressive power of LLMs. We provide both theoretical analysis and empirical evidence showing that WatME effectively preserves the diverse capabilities of LLMs while ensuring watermark detectability.