Large language models (LLMs) are increasingly impacting human society, particularly in textual information. Based on more than 30,000 papers and 1,000 presentations from machine learning conferences, we examined and compared the words used in writing and speaking, representing the first large-scale study of how LLMs influence the two main modes of verbal communication and expression within the same group of people. Our empirical results show that LLM-style words such as significant have been used more frequently in abstracts and oral presentations. The implicit impact on human expression like writing and speaking is beginning to emerge and is likely to grow in the future. We take the first step in building an automated monitoring platform to record its longitudinal changes to call attention to the implicit influence and ripple effect of LLMs on human society.
论文检索
输入标题、作者或关键词，从 3,086 篇学术成果中精准定位
Large Language Models as Neurolinguistic Subjects: Discrepancy between Performance and Competence
PDF ↗This study investigates the linguistic understanding of Large Language Models (LLMs) regarding signifier (form) and signified (meaning) by distinguishing two LLM assessment paradigms: psycholinguistic and neurolinguistic. Traditional psycholinguistic evaluations often reflect statistical rules that may not accurately represent LLMs’ true linguistic competence. We introduce a neurolinguistic approach, utilizing a novel method that combines minimal pair and diagnostic probing to analyze activation patterns across model layers. This method allows for a detailed examination of how LLMs represent form and meaning, and whether these representations are consistent across languages. We found: (1) Psycholinguistic and neurolinguistic methods reveal that language performance and competence are distinct; (2) Direct probability measurement may not accurately assess linguistic competence; (3) Instruction tuning won’t change much competence but improve performance; (4) LLMs exhibit higher competence and performance in form compared to meaning. Additionally, we introduce new conceptual minimal pair datasets for Chinese (COMPS-ZH) and German (COMPS-DE), complementing existing English datasets.
Human activity is moderated by norms; however, supervision for normative reasoning is sparse, particularly where norms are physically- or socially-grounded. We thus present EgoNormia \lVert \epsilon \rVert, comprising 1,853 (200 for EgoNormia-verified) multiple choice questions (MCQs) grounded within ego-centric videos of human interactions, enabling the evaluation and improvement of normative reasoning in vision-language models (VLMs). spans seven norm categories: safety, privacy, proxemics, politeness, cooperation, coordination/proactivity, and communication/legibility. To compile this dataset at scale, we propose a novel pipeline to generate grounded MCQs from raw egocentric video. Our work demonstrates that current state-of-the-art VLMs lack robust grounded norm understanding, scoring a maximum of 54% on EgoNormia and 58% on EgoNormia-verified, with performance across norm categories indicating significant risks of safety and privacy when VLMs are used in real-world agents. We additionally explore methods for improving normative understanding, demonstrating a naive retrieval-based generation (RAG) method using can enhance normative reasoning in VLMs.
Recent advances in large language models have led to numerous task-specialized fine-tuned variants, creating a need for efficient model merging techniques that preserve specialized capabilities while avoiding costly retraining. While existing task vector-based merging methods show promise, they typically apply uniform coefficients across all parameters, overlooking varying parameter importance both within and across tasks. We present Sens-Merging, a sensitivity-guided coefficient adjustment method that enhances existing model merging techniques by operating at both task-specific and cross-task levels. Our method analyzes parameter sensitivity within individual tasks and evaluates cross-task transferability to determine optimal merging coefficients. Extensive experiments on Mistral 7B and LLaMA2 7B/13B models demonstrate that Sens-Merging significantly improves performance across general knowledge, mathematical reasoning, and code generation tasks. Notably, when combined with existing merging techniques, our method enables merged models to outperform specialized fine-tuned models, particularly in code generation tasks. Our findings reveal important trade-offs between task-specific and cross-task scalings, providing insights for future model merging strategies.
This study intends to systematically disentangle pure logic reasoning and text understanding by investigating the contrast across abstract and contextualized logical problems from a comprehensive set of domains. We explore whether LLMs demonstrate genuine reasoning capabilities across various domains when the underlying logical structure remains constant. We focus on two main questions (1) Can abstract logical problems alone accurately benchmark LLMs’ reasoning ability in real-world scenarios, disentangled from contextual support in practical settings? (2) Does fine-tuning LLMs on abstract logic problems generalize to contextualized logic problems and vice versa? To investigate these questions, we focus on standard propositional logic, specifically propositional deductive and abductive logic reasoning. We construct datasets for both reasoning types with four difficulty levels across 12 distinct domains based on the Wikipedia categorization in addition to those with purely abstract variables. Our experiments aim to provide insights into disentangling context in logical reasoning, the genuine reasoning capabilities of LLMs, and their generalization potential. Coda and data are available at https://anonymous.4open.science/r/ContextHub-957E.
Text-to-SQL demands precise reasoning to convert natural language questions into structured queries. While large language models (LLMs) excel in many reasoning tasks, their ability to leverage Chain-of-Thought (CoT) reasoning for text-to-SQL remains underexplored. We identify critical limitations: zero-shot CoT offers minimal gains, and Direct Preference Optimization (DPO) applied without CoT yields marginal improvements. We propose ExCoT-DPO, a novel framework that iteratively optimizes open-source LLMs by combining CoT reasoning with off-policy and on-policy DPO, relying solely on execution accuracy as feedback. This approach eliminates the need for reward models or human-annotated preferences. Our experimental results demonstrate significant performance gains: ExCoT-DPO improves execution accuracy on BIRD from 57.37% to 68.51% and on Spider from 78.81% to 86.59% for LLaMA-3 70B, with Qwen-2.5-Coder demonstrating similar improvements. Our best model achieves state-of-the-art performance in the single-model setting on both BIRD and Spider datasets.
Worse than Random? An Embarrassingly Simple Probing Evaluation of Large Multimodal Models in Medical VQA
PDF ↗Large Multimodal Models (LMMs) have demonstrated impressive performance on existing medical Visual Question Answering (Med-VQA) benchmarks. However, high reported accuracy does not necessarily reflect their true diagnostic reliability in clinical settings. This study reveals that state-of-the-art models perform worse than random guessing on medical diagnosis questions when subjected to simple Probing Evaluation for Medical Diagnosis (ProbMed). ProbMed challenges models through probing evaluation and procedural diagnosis. Particularly, probing evaluation features pairing ground-truth questions with adversarial counterparts that feature negated and hallucinated attributes, while procedural diagnosis requires reasoning across various dimensions for each image, including modality recognition, organ identification, clinical findings, abnormalities, and positional grounding. Our evaluation reveals that even top-performing models like GPT-4o, GPT-4V, and Gemini Pro perform worse than random guessing on specialized diagnostic questions, indicating significant limitations in handling fine-grained medical inquiries. Furthermore, our ablation study on open-source models (e.g., LLaVA, LLaVA-Med, and Med-Flamingo) identifies poor visual understanding as a primary bottleneck—a limitation that can be partially mitigated by incorporating visual descriptions generated by GPT-4o, resulting in an average performance improvement of 9.44%. These findings underscore the urgent need for more robust evaluation methods and domain-specific expertise to ensure the reliability of LMMs in high-stakes medical applications.
Query expansion methods powered by large language models (LLMs) have demonstrated effectiveness in zero-shot retrieval tasks. These methods assume that LLMs can generate hypothetical documents that, when incorporated into a query vector, enhance the retrieval of real evidence. However, we challenge this assumption by investigating whether knowledge leakage in benchmarks contributes to the observed performance gains. Using fact verification as a testbed, we analyze whether the generated documents contain information entailed by ground-truth evidence and assess their impact on performance. Our findings indicate that, on average, performance improvements consistently occurred for claims whose generated documents included sentences entailed by gold evidence. This suggests that knowledge leakage may be present in fact-verification benchmarks, potentially inflating the perceived performance of LLM-based query expansion methods.
From Observation to Understanding: Front-Door Adjustments with Uncertainty Calibration for Enhancing Egocentric Reasoning in LVLMs
PDF ↗Recent progress in large vision-language models (LVLMs) has shown substantial potential across a broad spectrum of third-person tasks. However, adapting these LVLMs to egocentric scenarios remains challenging due to their third-person training bias. Existing methods that adapt LVLMs for first-person tasks often overlook critical agent-environment interactions, limiting their ability to perform egocentric reasoning. To address these challenges, we propose a novel zero-shot paradigm termed Front-Door Adjustments with Uncertainty Calibration (FRUIT) to enhance the egocentric reasoning abilities of LVLMs by simulating human causal reasoning. Specifically, the FRUIT operates in two stages: observation and understanding. Unlike conventional prompting techniques, we formalize egocentric reasoning using a structural causal model. Then, we ground interaction regions and expand them into hierarchical visual cues, augmented with corresponding captions, to form the initial observations. To reduce noise in these observations, we employ uncertainty calibration to filter out unreliable information. These refined observations as mediators are then incorporated into the prompt template, guiding the model to understand semantics from a first-person perspective. Extensive experiments conducted on the EgoThink benchmark demonstrate that our FRUIT method consistently enhances the performance of existing LVLMs on six distinct tasks. Our code is available at https://github.com/Mrshenshen/FRUIT.
Charts are ubiquitous, as people often use them to analyze data, answer questions, and discover critical insights. However, performing complex analytical tasks with charts requires significant perceptual and cognitive effort. Chart Question Answering (CQA) systems automate this process by enabling models to interpret and reason with visual representations of data. However, existing benchmarks like ChartQA lack real-world diversity and have recently shown performance saturation with modern large vision-language models (LVLMs). To address these limitations, we introduce ChartQAPro, a new benchmark that includes 1,341 charts from 99 diverse sources, spanning various chart types—including infographics and dashboards—and featuring 1,948 questions in various types, such as multiple-choice, conversational, hypothetical, and unanswerable questions, to better reflect real-world challenges. Our evaluations with 21 models show a substantial performance drop for LVLMs on ChartQAPro; e.g., Claude Sonnet 3.5 scores 90.5% on ChartQA but only 55.81% on ChartQAPro, underscoring the complexity of chart reasoning. We complement our findings with detailed error analyses and ablation studies, identifying key challenges and opportunities for advancing LVLMs in chart understanding and reasoning. We release ChartQAPro at https://github.com/vis-nlp/ChartQAPro.
Large language models are trained on massive scrapes of the web, as required by current scaling laws. Most progress is made for English, given its abundance of high-quality pretraining data. For most other languages, however, such high quality pretraining data is unavailable. In this work, we study how to boost pretrained model performance in a target language with insufficient pretraining data for training a high performing language model by enlisting data from an auxiliary language for which high quality data is available. We study this by quantifying the performance gap between training with data in a data-rich auxiliary language compared with training in the target language, exploring the benefits of translation systems, studying the limitations of model scaling when data is limited in the target languages, and proposing new methods for upsampling data from the auxiliary language. Our results show that stronger auxiliary datasets result in performance gains without modification to the model or training objective for close languages, and, in particular, that performance gains due to the development of more information-rich English pretraining datasets can extend to targeted language settings with limited data.
Large-scale multilingual evaluations, such as MEGA, often include only a handful of African languages due to the scarcity of high-qualityevaluation data and the limited discoverability of existing African datasets. This lack of representation hinders comprehensive LLM evaluation across a diverse range of languages and tasks. To address these challenges, we introduce AFROBENCH—a multi-task benchmark for evaluating the performance of LLMs across 64 African languages, 15 tasks and 22 datasets. AFROBENCH consists of nine natural language understanding datasets, six text generation datasets, six knowledge and question answering tasks, and one mathematical reasoning task. We present results comparing the performance of prompting LLMs to fine-tuned baselines based on BERT and T5-style models. Our results suggest large gaps in performance between high-resource languages, such as English, and African languages across most tasks; but performance also varies based on the availability of monolingual data resources. Our findings confirm that performance on African languages continues to remain a hurdle for current LLMs, underscoring the need for additional efforts to close this gap.
V-ALPHASOCIAL: Benchmark and Self-Reflective Chain-of-Thought Generation for Visual Social Commonsense Reasoning
PDF ↗Social commonsense reasoning naturally involves both the verbal and non-verbal cues of a social interaction. It is important for Large Vision-Language Models (VLMs) to leverage both textual and visual information in performing tasks like social understanding and reasoning. However, while current LLMs have shown good social reasoning capabilities in textual context, whether they can effectively incorporate visual information in social comprehension remains under-explored. To narrow the gap, we first construct and propose a benchmark: V-Social, featuring well-aligned text and visual content, tailored to assess visual social commonsense for multimodal foundation models. Through experimenting with V-Social, we find that even the most advanced VLM, GPT-4o, often falls short in social commonsense reasoning. This highlights the critical need to enhance the social grounding of VLMs. One major obstacle for improving this is the lack of high-quality data with good reasoning process. To overcome this obstacle, we introduce V-AlphaSocial, a novel method that generates high-quality chain-of-thought reasoning paths from unlabeled data. We design a visual reasoning reward model to improve VLM, and then iteratively refine both the VLM and the reward model. Our extensive analysis showcases how our method enhances social commonsense reasoning, proposing an effective approach that facilitates deeper exploration into field.
In retrieval augmented generation (RAG) systems, each individual component—the LLM, embedder, and corpus—could introduce biases in the form of skews towards certain genders or political leanings. In this work, we study the conflict between biases of each component and their relationship to the overall bias of the RAG system, which we call \textit{bias conflict}. Examining both gender and political biases as case studies, we show that bias conflict can be characterized through a linear relationship among components despite its complexity. Through fine-tuning, we demonstrate how to control the bias of the embedder while maintaining utility and reveal the importance of \textit{reverse-biasing} the embedder to mitigate bias in the overall system, Additionally, we find that LLMs and tasks exhibit varying \textit{sensitivities} to bias, a crucial factor to consider for debiasing. Our results underscore that a fair RAG system can be better achieved by carefully controlling the bias of the embedder rather than increasing its fairness.
Large language models (LLMs) exhibit pronounced conservative bias in relation extraction tasks, frequently defaulting to no_relation label when an appropriate option is unavailable. While this behavior helps prevent incorrect relation assignments, our analysis reveals that it also leads to significant information loss when reasoning is not explicitly included in the output. We systematically evaluate this trade-off across multiple prompts, datasets, and relation types, introducing the concept of Hobson’s choice to capture scenarios where models opt for safe but uninformative labels over hallucinated ones. Our findings suggest that conservative bias occurs twice as often as hallucination. To quantify this effect, we use SBERT and LLM prompts to capture the semantic similarity between conservative bias behaviors in constrained prompts and labels generated from semi-constrained and open-ended prompts.
TReMu: Towards Neuro-Symbolic Temporal Reasoning for LLM-Agents with Memory in Multi-Session Dialogues
PDF ↗Temporal reasoning in multi-session dialogues presents a significant challenge which has been under-studied in previous temporal reasoning benchmarks. To bridge this gap, we propose a new evaluation task for temporal reasoning in multi-session dialogues and introduce an approach to construct a new benchmark by augmenting dialogues from LoCoMo and creating multi-choice QAs. Furthermore, we present TReMu, a new framework aimed at enhancing the temporal reasoning capabilities of LLM-agents in this context. Specifically, the framework employs time-aware memorization through timeline summarization, generating retrievable memory by summarizing events in each dialogue session with their inferred dates. Additionally, we integrate neuro-symbolic temporal reasoning, where LLMs generate Python code to perform temporal calculations and select answers. Experimental evaluations on popular LLMs demonstrate that our benchmark is challenging, and the proposed framework significantly improves temporal reasoning performance compared to baseline methods, raising from 29.83 on GPT-4o via standard prompting to 77.67 via our approach and highlighting its effectiveness in addressing temporal reasoning in multi-session dialogues.
Systematic Evaluation of Auto-Encoding and Large Language Model Representations for Capturing Author States and Traits
PDF ↗Large Language Models (LLMs) are increasingly used in human-centered applications, yet their ability to model diverse psychological constructs is not well understood. In this study, we systematically evaluate a range of Transformer-LMs to predict psychological variables across five major dimensions: affect, substance use, mental health, sociodemographics, and personality. Analyses span three temporal levels—short daily text responses about current affect, text aggregated over two-weeks, and user-level text collected over two years—allowing us to examine how each model’s strengths align with the underlying stability of different constructs. The findings show that mental health signals emerge as the most accurately predicted dimensions (r=0.6) across all temporal scales. At the daily scale, smaller models like DeBERTa and HaRT often performed better, whereas, at longer scales or with greater context, larger model like Llama3-8B performed the best. Also, aggregating text over the entire study period yielded stronger correlations for outcomes, such as age and income. Overall, these results suggest the importance of selecting appropriate model architectures and temporal aggregation techniques based on the stability and nature of the target variable.
The lack of explainability in state-of-the-art Natural Language Understanding (NLU) classification models has increased interest in developing techniques for improving explainable linear feature-based models (e.g., Logistic Regression/SVM). Politeness detection is a task that exemplifies this interest. While those techniques perform well on the task when applied to data from the same domain as the training data, they lack generalizability and thus fall short when applied to data from other domains. This is due to their reliance on discovering domain-specific word-level features. We introduce a method for improving the generalizability of explainable politeness models by relying on speech act patterns instead of words, leveraging speech act labels assigned by the GPT-4 model. This approach goes beyond the mere words and injects intent into politeness classification models, enhancing their generalizability. Results demonstrate that the proposed method achieves state-of-the-art accuracy in the cross-domain setting among explainable methods, while falling short in the in-domain setting. Our findings illustrate that explainable models can benefit from Large Language Models.
The Lies Characters Tell: Utilizing Large Language Models to Normalize Adversarial Unicode Perturbations
PDF ↗Homoglyphs, Unicode characters that are visually homogeneous to Latin letters, are widely used to mask offensive content. Dynamic strategies are needed to combat homoglyphs as the Unicode library is ever-expanding and new substitution possibilities for Latin letters continuously emerge. The present study investigated two novel mitigation approaches that do not rely on strict mappings but instead harness the power of large language models to neutralize both known and unknown homoglyphs: (1) indirectly normalizing homoglyphs by replacing non-Latin characters with a delimiter and prompting large language models to “fill in the blanks” and (2) directly normalizing homoglyphs by using large language models to determine which characters should be replaced with Latin letters. We found that GPT-4o-mini constructed normalized text with an average cosine similarity score of 0.91 to the original tweets when applying our indirect method and 0.96 to the original tweets when applying our direct method. This study indicates that large language model-based normalization techniques can effectively unmask offensive content concealed by homoglyphs. Code and data are available in our GitHub repository: https://github.com/pcoopercoder/The-Lies-Characters-Tell.
Let The Jury Decide: Fair Demonstration Selection for In-Context Learning through Incremental Greedy Evaluation
PDF ↗Large Language Models (LLMs) are powerful in-context learners, achieving strong performance with just a few high-quality demonstrations. However, fairness concerns arise in many in-context classification tasks, especially when predictions involve sensitive attributes. To address this, we propose JUDGE—a simple yet effective framework for selecting fair and representative demonstrations that improve group fairness in In-Context Learning. JUDGE constructs the demonstration set iteratively using a greedy approach, guided by a small, carefully selected jury set. Our method remains robust across varying LLM architectures and datasets, ensuring consistent fairness improvements. We evaluate JUDGE on four datasets using four LLMs, comparing it against seven baselines. Results show that JUDGE consistently improves fairness metrics without compromising accuracy.