论文检索

Xiaolong Wang, Yile Wang, Yuanchi Zhang, Fuwen Luo, Peng Li, Maosong Sun, Yang Liu

Large Language Models (LLMs) have achieved remarkable performance in objective tasks such as open-domain question answering and mathematical reasoning, which can often be solved through recalling learned factual knowledge or chain-of-thought style reasoning. However, we find that the performance of LLMs in subjective tasks is still unsatisfactory, such as metaphor recognition, dark humor detection, etc. Compared to objective tasks, subjective tasks focus more on interpretation or emotional response rather than a universally accepted reasoning pathway. Based on the characteristics of the tasks and the strong dialogue-generation capabilities of LLMs, we propose RiC (Reasoning in Conversation), a method that focuses on solving subjective tasks through dialogue simulation. The motivation of RiC is to mine useful contextual information by simulating dialogues instead of supplying chain-of-thought style rationales, thereby offering potential useful knowledge behind dialogues for giving the final answers. We evaluate both API-based and open-source LLMs including GPT-4, ChatGPT, and OpenChat across twelve tasks. Experimental results show that RiC can yield significant improvement compared with various baselines.

Nina Rimsky, Nick Gabrieli, Julian Schulz, Meg Tong, Evan Hubinger, Alexander Turner

We introduce Contrastive Activation Addition (CAA), a method for steering language models by modifying their activations during forward passes. CAA computes “steering vectors” by averaging the difference in residual stream activations between pairs of positive and negative examples of a particular behavior, such as factual versus hallucinatory responses. During inference, these steering vectors are added at all token positions after the user’s prompt with either a positive or negative coefficient, allowing precise control over the degree of the targeted behavior. We evaluate CAA’s effectiveness on Llama 2 Chat using multiple-choice behavioral question datasets and open-ended generation tasks. We demonstrate that CAA significantly alters model behavior, is effective over and on top of traditional methods like finetuning and system prompt design, and minimally reduces capabilities. Moreover, we gain deeper insights into CAA’s mechanisms by employing various activation space interpretation methods. CAA accurately steers model outputs and sheds light on how high-level concepts are represented in Large Language Models (LLMs).

Yulia Otmakhova, Shima Khanehzar, Lea Frermann

Framing studies how individuals and societies make sense of the world, by communicating or representing complex issues through schema of interpretation. The framing of information in the mass media influences our interpretation of facts and corresponding decisions, so detecting and analysing it is essential to understand biases in the information we consume. Despite that, framing is still mostly examined manually, on a case-by-case basis, while existing large-scale automatic analyses using NLP methods are not mature enough to solve this task. In this survey we show that despite the growing interest to framing in NLP its current approaches do not capture those aspects which allow to frame, rather than simply convey, the message. To this end, we bring together definitions of frames and framing adopted in different disciplines; examine cognitive, linguistic, and communicative aspects a frame contains beyond its topical content. We survey recent work on computational frame detection, and discuss how framing aspects and frame definitions are (or should) be reflected in NLP approaches.

Fengqing Jiang, Zhangchen Xu, Luyao Niu, Zhen Xiang, Bhaskar Ramasubramanian, Bo Li, Radha Poovendran

Safety is critical to the usage of large language models (LLMs). Multiple techniques such as data filtering and supervised fine-tuning have been developed to strengthen LLM safety. However, currently known techniques presume that corpora used for safety alignment of LLMs are solely interpreted by semantics. This assumption, however, does not hold in real-world applications, which leads to severe vulnerabilities in LLMs. For example, users of forums often use ASCII art, a form of text-based art, to convey image information. In this paper, we propose a novel ASCII art-based jailbreak attack and introduce a comprehensive benchmark Vision-in-Text Challenge (ViTC) to evaluate the capabilities of LLMs in recognizing prompts that cannot be solely interpreted by semantics. We show that five SOTA LLMs (GPT-3.5, GPT-4, Gemini, Claude, and Llama2) struggle to recognize prompts provided in the form of ASCII art. Based on this observation, we develop the jailbreak attack ArtPrompt, which leverages the poor performance of LLMs in recognizing ASCII art to bypass safety measures and elicit undesired behaviors from LLMs. ArtPrompt only requires black-box access to the victim LLMs, making it a practical attack. We evaluate ArtPrompt on five SOTA LLMs, and show that ArtPrompt can effectively and efficiently induce undesired behaviors from all five LLMs.

Haoqiu Yan, Yongxin Zhu, Kai Zheng, Bing Liu, Haoyu Cao, Deqiang Jiang, Linli Xu

Large Language Model (LLM)-enhanced agents become increasingly prevalent in Human-AI communication, offering vast potential from entertainment to professional domains. However, current multi-modal dialogue systems overlook the acoustic information present in speech, which is crucial for understanding human communication nuances. This oversight can lead to misinterpretations of speakers’ intentions, resulting in inconsistent or even contradictory responses within dialogues. To bridge this gap, in this paper, we propose PerceptiveAgent, an empathetic multi-modal dialogue system designed to discern deeper or more subtle meanings beyond the literal interpretations of words through the integration of speech modality perception. Employing LLMs as a cognitive core, PerceptiveAgent perceives acoustic information from input speech and generates empathetic responses based on speaking styles described in natural language. Experimental results indicate that PerceptiveAgent excels in contextual understanding by accurately discerning the speakers’ true intentions in scenarios where the linguistic meaning is either contrary to or inconsistent with the speaker’s true feelings, producing more nuanced and expressive spoken dialogues. Code is publicly available at: https://github.com/Haoqiu-Yan/PerceptiveAgent.

Aryaman Arora, Dan Jurafsky, Christopher Potts

Language models (LMs) have proven to be powerful tools for psycholinguistic research, but most prior work has focused on purely behavioural measures (e.g., surprisal comparisons). At the same time, research in model interpretability has begun to illuminate the abstract causal mechanisms shaping LM behavior. To help bring these strands of research closer together, we introduce CausalGym. We adapt and expand the SyntaxGym suite of tasks to benchmark the ability of interpretability methods to causally affect model behaviour. To illustrate how CausalGym can be used, we study the pythia models (14M–6.9B) and assess the causal efficacy of a wide range of interpretability methods, including linear probing and distributed alignment search (DAS). We find that DAS outperforms the other methods, and so we use it to study the learning trajectory of two difficult linguistic phenomena in pythia-1b: negative polarity item licensing and filler–gap dependencies. Our analysis shows that the mechanism implementing both of these tasks is learned in discrete stages, not gradually.

Lei Li, Yuqi Wang, Runxin Xu, Peiyi Wang, Xiachong Feng, Lingpeng Kong, Qi Liu

Large vision-language models (LVLMs) excel across diverse tasks involving concrete images from natural scenes. However, their ability to interpret abstract figures, such as geometry shapes and scientific plots, remains limited due to a scarcity of training datasets in scientific domains.To fill this gap, we introduce Multimodal ArXiv, consisting of ArXivCap and ArXivQA, for enhancing LVLMs scientific comprehension.ArXivCap is a figure-caption dataset comprising 6.4M images and 3.9M captions, sourced from 572K ArXiv papers spanning various scientific domains.Drawing from ArXivCap, we introduce ArXivQA, a question-answering dataset generated by prompting GPT-4V based on scientific figures. ArXivQA greatly enhances open-sourced LVLMs’ mathematical reasoning capabilities, achieving a 10.4% absolute accuracy gain on a multimodal mathematical reasoning benchmark.Furthermore, employing ArXivCap, we devise four vision-to-text tasks for benchmarking LVLMs.Evaluation results with state-of-the-art LVLMs underscore their struggle with the nuanced semantics of academic figures, while domain-specific training yields substantial performance gains.Our error analysis uncovers misinterpretations of visual context, recognition errors, and the production of overly simplified captions by current LVLMs, shedding light on future improvements.

Jinghan Zhang, Xiting Wang, Yiqiao Jin, Changyu Chen, Xinhao Zhang, Kunpeng Liu

The reward model for Reinforcement Learning from Human Feedback (RLHF) has proven effective in fine-tuning Large Language Models (LLMs). Notably, collecting human feedback for RLHF can be resource-intensive and lead to scalability issues for LLMs and complex tasks. Our proposed framework Proto-RM leverages prototypical networks to enhance reward models under limited human feedback. By enabling stable and reliable structural learning from fewer samples, Proto-RM significantly enhances LLMs' adaptability and accuracy in interpreting human preferences. Extensive experiments on various datasets demonstrate that Proto-RM significantly improves the performance of reward models and LLMs in human feedback tasks, achieving comparable and usually better results than traditional methods, while requiring significantly less data in data-limited scenarios. This research offers a promising direction for enhancing the efficiency of reward models and optimizing the fine-tuning of language models under restricted feedback conditions.

Tenggan Zhang, Xinjie Zhang, Jinming Zhao, Li Zhou, Qin Jin

Understanding the reason for emotional support response is crucial for establishing connections between users and emotional support dialogue systems. Previous works mostly focus on generating better responses but ignore interpretability, which is extremely important for constructing reliable dialogue systems. To empower the system with better interpretability, we propose an emotional support response generation scheme, named \textbf{E}motion-Focused and \textbf{S}trategy-Driven \textbf{C}hain-\textbf{o}f-\textbf{T}hought (\textbf{ESCoT}), mimicking the process of \textit{identifying}, \textit{understanding}, and \textit{regulating} emotions. Specially, we construct a new dataset with ESCoT in two steps: (1) \textit{Dialogue Generation} where we first generate diverse conversation situations, then enhance dialogue generation using richer emotional support strategies based on these situations; (2) \textit{Chain Supplement} where we focus on supplementing selected dialogues with elements such as emotion, stimuli, appraisal, and strategy reason, forming the manually verified chains. Additionally, we further develop a model to generate dialogue responses with better interpretability. We also conduct extensive experiments and human evaluations to validate the effectiveness of the proposed ESCoT and generated dialogue responses. Our dataset, code, and model will be released.

Yilun Zhao, Lyuhao Chen, Arman Cohan, Chen Zhao

Long-form Table Question Answering (LFTQA) requires systems to generate paragraph long and complex answers to questions over tabular data. While Large language models based systems have made significant progress, it often hallucinates, especially when the task involves complex reasoning over tables. To tackle this issue, we propose a new LLM-based framework, TaPERA, for LFTQA tasks. Our framework uses a modular approach that decomposes the whole process into three sub-modules: 1) QA-based Content Planner that iteratively decomposes the input question into sub-questions; 2) Execution-based Table Reasoner that produces executable Python program for each sub-question; and 3) Answer Generator that generates long-form answer grounded on the program output. Human evaluation results on the FeTaQA and QTSumm datasets indicate that our framework significantly improves strong baselines on both accuracy and truthfulness, as our modular framework is better at table reasoning, and the long-form answer is always consistent with the program output. Our modular design further provides transparency as users are able to interact with our framework by manually changing the content plans.

Giulia Rambelli, Emmanuele Chersoni, Claudia Collacciani, Marianna Bolognesi

Noun-noun compounds interpretation is the task where a model is given one of such constructions, and it is asked to provide a paraphrase, making the semantic relation between the nouns explicit, as in carrot cake is “a cake made of carrots.” Such a task requires the ability to understand the implicit structured representation of the compound meaning. In this paper, we test to what extent the recent Large Language Models can interpret the semantic relation between the constituents of lexicalized English compounds and whether they can abstract from such semantic knowledge to predict the semantic relation between the constituents of similar but novel compounds by relying on analogical comparisons (e.g., carrot dessert). We test both Surprisal metrics and prompt-based methods to see whether i.) they can correctly predict the relation between constituents, and ii.) the semantic representation of the relation is robust to paraphrasing. Using a dataset of lexicalized and annotated noun-noun compounds, we find that LLMs can infer some semantic relations better than others (with a preference for compounds involving concrete concepts). When challenged to perform abstractions and transfer their interpretations to semantically similar but novel compounds, LLMs show serious limitations.

Claudia Collacciani, Giulia Rambelli, Marianna Bolognesi

Generics are expressions used to communicate abstractions about categories. While conveying general truths (e.g., “Birds fly”), generics have the interesting property to admit exceptions (e.g., penguins do not fly). Statements of this type help us organizing our knowledge of the world, and form the basis of how we express it (Hampton, 2012; Leslie, 2014).This study investigates how Large Language Models (LLMs) interpret generics, drawing upon psycholinguistic experimental methodologies. Understanding how LLMs interpret generic statements serves not only as a measure of their ability to abstract but also arguably plays a role in their encoding of stereotypes. Given that generics interpretation necessitates a comparison with explicitly quantified sentences, we explored i.) whether LLMs can correctly associate a quantifier with the generic structure, and ii.) whether the presence of a generic sentence as context influences the outcomes of quantifiers. We evaluated LLMs using both Surprisal distributions and prompting techniques.The findings indicate that models do not exhibit a strong sensitivity to quantification. Nevertheless, they seem to encode a meaning linked with the generic structure, which leads them to adjust their answers accordingly when a generalization is provided as context.

Xiao Liu, Xuanyu Lei, Shengyuan Wang, Yue Huang, Andrew Feng, Bosi Wen, Jiale Cheng, Pei Ke, Yifan Xu, Weng Lam Tam 等

Alignment has become a critical step for instruction-tuned Large Language Models (LLMs) to become helpful assistants. However, effective evaluation of alignment for emerging Chinese LLMs is still significantly lacking, calling for real-scenario grounded, open-ended, challenging and automatic evaluations tailored for alignment. To fill in this gap, we introduce AlignBench, a comprehensive multi-dimensional benchmark for evaluating LLMs’ alignment in Chinese. We tailor a human-in-the-loop data curation pipeline, containing 8 main categories, 683 real-scenario rooted queries and corresponding human verified references.To ensure references’ correctness, each knowledge-intensive query is accompanied with evidences collected from reliable webpages (including the url and quotation) by our annotators.For automatic evaluation, our benchmark employs a rule-calibrated multi-dimensional LLM-as-Judge (CITATION) with Chain-of-Thought to generate explanations and final ratings as evaluations, ensuring high reliability and interpretability.All evaluation codes and data are publicly available at https://github.com/THUDM/AlignBench

Anwoy Chatterjee, Eshaan Tanwar, Subhabrata Dutta, Tanmoy Chakraborty

Large Language Models (LLMs) have transformed NLP with their remarkable In-context Learning (ICL) capabilities. Automated assistants based on LLMs are gaining popularity; however, adapting them to novel tasks is still challenging. While colossal models excel in zero-shot performance, their computational demands limit widespread use, and smaller language models struggle without context. This paper investigates whether LLMs can generalize from labeled examples of predefined tasks to novel tasks. Drawing inspiration from biological neurons and the mechanistic interpretation of the Transformer architecture, we explore the potential for information sharing across tasks. We design a cross-task prompting setup with three LLMs and show that LLMs achieve significant performance improvements despite no examples from the target task in the context. Cross-task prompting leads to a remarkable performance boost of 107% for LLaMA-2 7B, 18.6% for LLaMA-2 13B, and 3.2% for GPT 3.5 on average over zero-shot prompting, and performs comparable to standard in-context learning. The effectiveness of generating pseudo-labels for in-task examples is demonstrated, and our analyses reveal a strong correlation between the effect of cross-task examples and model activation similarities in source and target input tokens. This paper offers a first-of-its-kind exploration of LLMs’ ability to solve novel tasks based on contextual signals from different task examples.

Changjiang Gao, Jixing Li, Jiajun Chen, Shujian Huang

The process of meaning composition, wherein smaller units like morphemes or words combine to form the meaning of phrases and sentences, is essential for human sentence comprehension. Despite extensive neurolinguistic research into the brain regions involved in meaning composition, a computational metric to quantify the extent of composition is still lacking. Drawing on the key-value memory interpretation of transformer feed-forward network blocks, we introduce the Composition Score, a novel model-based metric designed to quantify the degree of meaning composition during sentence comprehension. Experimental findings show that this metric correlates with brain clusters associated with word frequency, structural processing, and general sensitivity to words, suggesting the multifaceted nature of meaning composition during human sentence comprehension.

Ying-Chun Lin, Jennifer Neville, Jack Stokes, Longqi Yang, Tara Safavi, Mengting Wan, Scott Counts, Siddharth Suri, Reid Andersen, Xiaofeng Xu 等

Accurate and interpretable user satisfaction estimation (USE) is critical for understanding, evaluating, and continuously improving conversational systems. Users express their satisfaction or dissatisfaction with diverse conversational patterns in both general-purpose (ChatGPT and Bing Copilot) and task-oriented (customer service chatbot) conversational systems. Existing approaches based on featurized ML models or text embeddings fall short in extracting generalizable patterns and are hard to interpret. In this work, we show that LLMs can extract interpretable signals of user satisfaction from their natural language utterances more effectively than embedding-based approaches. Moreover, an LLM can be tailored for USE via an iterative prompting framework using supervision from labeled examples. Our proposed method, Supervised Prompting for User satisfaction Rubrics (SPUR), not only has higher accuracy but is more interpretable as it scores user satisfaction via learned rubrics with a detailed breakdown.

Rikui Huang, Wei Wei, Xiaoye Qu, Shengzhe Zhang, Dangyang Chen, Yu Cheng

Recently, Temporal Knowledge Graph Forecasting (TKGF) has emerged as a pivotal domain for forecasting future events. Unlike black-box neural network methods, rule-based approaches are lauded for their efficiency and interpretability. For this line of work, it is crucial to correctly estimate the predictive effectiveness of the rules, i.e., the confidence. However, the existing literature lacks in-depth investigation into how confidence evolves with time. Moreover, inaccurate and heuristic confidence estimation limits the performance of rule-based methods. To alleviate such issues, we propose a framework named TempValid to explicitly model the temporal validity of rules for TKGF. Specifically, we design a time function to model the interaction between temporal information with confidence. TempValid conceptualizes confidence and other coefficients as learnable parameters to avoid inaccurate estimation and combinatorial explosion. Furthermore, we introduce a rule-adversarial negative sampling and a time-aware negative sampling strategies to facilitate TempValid learning. Extensive experiments show that TempValid significantly outperforms previous state-of-the-art (SOTA) rule-based methods on six TKGF datasets. Moreover, it exhibits substantial advancements in cross-domain and resource-constrained rule learning scenarios.

Fuwen Luo, Chi Chen, Zihao Wan, Zhaolu Kang, Qidong Yan, Yingjie Li, Xiaolong Wang, Siyu Wang, Ziyue Wang, Xiaoyue Mi 等

Multimodal large language models (MLLMs) have demonstrated promising results in a variety of tasks that combine vision and language. As these models become more integral to research and applications, conducting comprehensive evaluations of their capabilities has grown increasingly important. However, most existing benchmarks fail to consider that, in certain situations, images need to be interpreted within a broader context. In this work, we introduce a new benchmark, named as CODIS, designed to assess the ability of models to use context provided in free-form text to enhance visual comprehension. Our findings indicate that MLLMs consistently fall short of human performance on this benchmark. Further analysis confirms that these models struggle to effectively extract and utilize contextual information to improve their understanding of images. This underscores the pressing need to enhance the ability of MLLMs to comprehend visuals in a context-dependent manner.

Dun-Ming Huang, Pol Van Rijn, Ilia Sucholutsky, Raja Marjieh, Nori Jacoby

Conversational tones — the manners and attitudes in which speakers communicate — are essential to effective communication. As Large Language Models (LLMs) become increasingly popular, it is necessary to characterize the divergences in their conversational tones relative to humans. Prior research relied on pre-existing taxonomies or text corpora, which suffer from experimenter bias and may not be representative of real-world distributions. Inspired by methods from cognitive science, we propose an iterative method for simultaneously eliciting conversational tones and sentences, where participants alternate between two tasks: (1) one participant identifies the tone of a given sentence and (2) a different participant generates a sentence based on that tone. We run 50 iterations of this process with both human participants and GPT-4 and obtain a dataset of sentences and frequent conversational tones. In an additional experiment, humans and GPT-4 annotated all sentences with all tones. With data from 1,339 participants, 33,370 human judgments, and 29,900 GPT-4 queries, we show how our approach can be used to create an interpretable geometric representation of relations between tones in humans and GPT-4. This work showcases how combining ideas from machine learning and cognitive science can address challenges in human-computer interactions.

Daniel Braun, Florian Matthes

Legal tasks and datasets are often used as benchmarks for the capabilities of language models. However, openly available annotated datasets are rare. In this paper, we introduce AGB-DE, a corpus of 3,764 clauses from German consumer contracts that have been annotated and legally assessed by legal experts. Together with the data, we present a first baseline for the task of detecting potentially void clauses, comparing the performance of an SVM baseline with three fine-tuned open language models and the performance of GPT-3.5. Our results show the challenging nature of the task, with no approach exceeding an F1-score of 0.54. While the fine-tuned models often performed better with regard to precision, GPT-3.5 outperformed the other approaches with regard to recall. An analysis of the errors indicates that one of the main challenges could be the correct interpretation of complex clauses, rather than the decision boundaries of what is permissible and what is not.