Large language models (LLMs) have advanced conversational AI assistants. However, systematically evaluating how well these assistants apply personalization—adapting to individual user preferences while completing tasks—remains challenging. Existing personalization benchmarks focus on chit-chat, non-conversational tasks, or narrow domains, failing to capture the complexities of personalized task-oriented assistance. To address this, we introduce PersonaLens, a comprehensive benchmark for evaluating personalization in task-oriented AI assistants. Our benchmark features diverse user profiles equipped with rich preferences and interaction histories, along with two specialized LLM-based agents: a user agent that engages in realistic task-oriented dialogues with AI assistants, and a judge agent that employs the LLM-as-a-Judge paradigm to assess personalization, response quality, and task success. Through extensive experiments with current LLM assistants across diverse tasks, we reveal significant variability in their personalization capabilities, providing crucial insights for advancing conversational AI systems.
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
Exploring Supervised Approaches to the Detection of Anthropomorphic Language in the Reporting of NLP Venues
PDF ↗We investigate the prevalence of anthropomorphic language in the reporting of AI technology, focussed on NLP and LLMs. We undertake a corpus annotation focussing on one year of ACL long-paper abstracts and news articles from the same period. We find that 74% of ACL abstracts and 88% of news articles contain some form of anthropomorphic description of AI technology. Further, we train a regression classifier based on BERT, demonstrating that we can automatically label abstracts for their degree of anthropomorphism based on our corpus. We conclude by applying this labelling process to abstracts available in the entire history of the ACL Anthology and reporting on diachronic and inter-venue findings, showing that the degree of anthropomorphism is increasing at all examined venues over time.
Significant efforts have been made in the NLP community to facilitate the automatic analysis of climate-related corpora by tasks such as climate-related topic detection, climate risk classification, question answering over climate topics, and many more. In this work, we perform a reproducibility study on 8 tasks and 29 datasets, testing 6 models. We find that many tasks rely heavily on surface-level keyword patterns rather than deeper semantic or contextual understanding. Moreover, we find that 96% of the datasets contain annotation issues, with 16.6% of the sampled wrong predictions of a zero-shot classifier being actually clear annotation mistakes, and 38.8% being ambiguous examples.These results call into question the reliability of current benchmarks to meaningfully compare models and highlight the need for improved annotation practices. We conclude by outlining actionable recommendations to enhance dataset quality and evaluation robustness.
Authorship Attribution (AA) seeks to identify the author of a given text, yet existing methods often struggle with trustworthiness and interpretability, particularly across different domains, languages, and stylistic variations. These challenges arise from the absence of uncertainty quantification and the inability of current models to adapt to diverse authorship tasks. To address these limitations, we introduce BEDAA, a Bayesian-Enhanced DeBERTa framework that integrates Bayesian reasoning with transformer-based language models to enable uncertainty-aware and interpretable authorship attribution. BEDAA achieves up to 19.69% improvement in F1-score across multiple authorship attribution tasks, including binary, multiclass, and dynamic authorship detection. By incorporating confidence ranking, uncertainty decomposition, and probabilistic reasoning, BEDAA improves robustness while offering transparent decision-making processes. Furthermore, BEDAA extends beyond traditional AA by demonstrating its effectiveness in human vs. machine-generated text classification, code authorship detection, and cross-lingual attribution. These advances establish BEDAA as a generalised, interpretable, and adaptable framework for modern authorship attribution challenges.
Knowledge distillation (KD) is a key technique for compressing large language models into smaller ones while preserving performance. Despite the recent traction of KD research, its effectiveness for smaller language models (LMs) and the mechanisms driving knowledge transfer remain underexplored. In this work, we present the first large-scale empirical and statistical analysis of KD across models ranging from 0.5B to 7B parameters on 14 complex reasoning tasks in a zero-shot setting. Our findings reveal that KD can improve the average performance of smaller models by up to 10%, with a peak task specific gain of 22%, while providing only marginal benefits (\sim 1.3%) for larger models. Surprisingly, teacher performance has a minimal impact on student outcomes, while teacher task expertise impacts KD effectiveness. A correlation study indicates that smaller LMs benefit more from KD, whereas larger LMs show diminished gains. Additionally, we uncover a misalignment between improvements in student performance and reasoning fidelity, suggesting that while KD enhances accuracy, it does not always maintain the structured decision-making processes of the teacher. Our ablation study further highlights the importance of teacher signals and logit smoothing in influencing students’ performance after distillation. Overall, our study offers a comprehensive empirical and statistical assessment of KD, highlighting both its benefits and trade-offs when distilling knowledge from larger to smaller LMs.
Annotating the Annotators: Analysis, Insights and Modelling from an Annotation Campaign on Persuasion Techniques Detection
PDF ↗Persuasion (or propaganda) techniques detection is a relatively novel task in Natural Language Processing (NLP). While there have already been a number of annotation campaigns, they have been based on heuristic guidelines, which have never been thoroughly discussed. Here, we present the first systematic analysis of a complex annotation task -detecting 22 persuasion techniques in memes-, for which we provided continuous expert oversight. The presence of an expert allowed us to critically analyze specific aspects of the annotation process. Among our findings, we show that inter-annotator agreement alone inadequately assessed annotation correctness. We thus define and track different error types, revealing that expert feedback shows varying effectiveness across error categories. This pattern suggests that distinct mechanisms underlie different kinds of misannotations. Based on our findings, we advocate for an expert oversight in annotation tasks and periodic quality audits. As an attempt to reduce the costs for this, we introduce a probabilistic model for optimizing intervention scheduling.
Despite recent advances in Video Large Language Models (VideoLLMs), effectively understanding long-form videos remains a significant challenge. Perceiving lengthy videos containing thousands of frames poses substantial computational burden. To mitigate this issue, this paper introduces Generative Frame Sampler (GenS), a plug-and-play module integrated with VideoLLMs to facilitate efficient lengthy video perception. Built upon a lightweight VideoLLM, GenS leverages its inherent vision-language capabilities to identify question-relevant frames. To facilitate effective retrieval, we construct GenS-Video-150K, a large-scale video instruction dataset with dense frame relevance annotations. Extensive experiments demonstrate that GenS consistently boosts the performance of various VideoLLMs, including open-source models (Qwen2-VL-7B, Aria-25B, LLaVA-Video-7B/72B) and proprietary assistants (GPT-4o, Gemini). When equipped with GenS, open-source VideoLLMs achieve impressive state-of-the-art results on long-form video benchmarks: LLaVA-Video-72B reaches 66.8 (+4.3) on LongVideoBench and 77.0 (+2.7) on MLVU, while Aria obtains 39.2 on HourVideo surpassing the Gemini-1.5-pro by 1.9 points.
JARVIS-VLA: Post-Training Large-Scale Vision Language Models to Play Visual Games with Keyboards and Mouse
PDF ↗Recently, action-based decision-making in open-world environments has gained significant attention. Visual Language Action (VLA) models, pretrained on large-scale web datasets, have shown promise in decision-making tasks. However, previous work has primarily focused on action post-training, often neglecting enhancements to the foundation model itself. In response, we introduce Act from Visual Language Post-Training (ActVLP), a novel training paradigm. ActVLP distinctively enhances the foundation model prior to action-specific tuning by first post-training it on a curated set of environment-specific visual and linguistic tasks using self-supervised learning. This initial stage significantly improves the model’s capabilities in world knowledge, visual recognition, and spatial grounding. Subsequently, this strengthened VLM undergoes action post-training via imitation learning on trajectory datasets.Following this paradigm, we develop JARVIS-VLA, the first VLA model in Minecraft that can follow human instructions on over 1k different atomic tasks, including crafting, smelting, cooking, mining, and killing. Our experiments demonstrate that our ActVLP paradigm leads to a significant 40% improvement over the best agent baseline on a diverse set of atomic tasks. Furthermore, JARVIS-VLA surpasses traditional imitation learning-based policies in Minecraft, achieving state-of-the-art performance. We have open-sourced the code, models, and datasets to foster further research.The project page can be found at https://craftjarvis.github.io/JarvisVLA.
Transparentize the Internal and External Knowledge Utilization in LLMs with Trustworthy Citation
PDF ↗While hallucinations of large language models could be alleviated through retrieval-augmented generation and citation generation, how the model utilizes internal knowledge is still opaque, and the trustworthiness of its generated answers remains questionable. In this work, we introduce Context-Prior Augmented Citation Generation task, requiring models to generate citations considering both external and internal knowledge while providing trustworthy references, with 5 evaluation metrics focusing on 3 aspects: answer helpfulness, citation faithfulness, and trustworthiness. We introduce RAEL, the paradigm for our task, and also design INTRALIGN, an integrated method containing customary data generation and an alignment algorithm. Our experimental results show that our method achieves a better cross-scenario performance with regard to other baselines. Our extended experiments further reveal that retrieval quality, question types, and model knowledge have considerable influence on the trustworthiness in citation generation.
Data visualizations, such as bar charts and histograms, are essential for analyzing and exploring data, enabling the effective communication of insights. While existing methods have been proposed to translate natural language descriptions into visualization queries, they focus solely on spoken languages, overlooking sign languages, which comprise about 200 variants used by 70 million Deaf and Hard-of-Hearing (DHH) individuals. To fill this gap, this paper proposes Sign2Vis, a sign language interface that enables the DHH community to engage more fully with data analysis. We first construct a paired dataset that includes sign language pose videos and their corresponding visualization queries. Using this dataset, we evaluate a variety of models, including both pipeline-based and end-to-end approaches. Extensive experiments, along with a user study involving 15 participants, demonstrate the effectiveness of Sign2Vis. Finally, we share key insights from our evaluation and highlight the need for more accessible and user-centered tools to support the DHH community in interactive data analytics.
C2KD: Cross-layer and Cross-head Knowledge Distillation for Small Language Model-based Recommendation
PDF ↗Sequential recommenders predict users’ next interactions based on historical behavior and are essential in modern recommendation systems. While Large Language Models (LLMs) show promise, their size and high inference costs limit deployment on resource-constrained devices. Small Language Models (SLMs) provide a more efficient alternative for edge devices, but bridging the recommendation performance gap between LLMs and SLMs remains challenging. Typical approaches like supervised fine-tuning or vanilla knowledge distillation (KD) often lead to suboptimal performance or even negative transfer. Our motivational experiments reveal key issues with vanilla KD methods: feature imitation suffers from redundancy and uneven recommendation ability across layers, while prediction mimicking faces conflicts caused by differing weight distributions of prediction heads. To address these challenges, we propose a simple yet effective framework, C2KD, to transfer task-relevant knowledge from two complementary dimensions. Specifically, our method incorporates: (1) cross-layer feature imitation, which uses a dynamic router to select the most relevant teacher layers and assimilate task-relevant knowledge from the teacher’s late layers, allowing the student to concentrate on the teacher’s specialized knowledge; and (2) cross-head logit distillation, which maps the intermediate features of the student to the teacher’s output head, thereby minimizing prediction discrepancies between the teacher and the student. Extensive experiments across diverse model families demonstrate that our approach enables 1B-parameter SLMs to achieve competitive performance compared to LLMs (e.g., Llama3-8B), offering a practical solution for real-world on-device sequential recommendations.
This paper introduces PeopleJoin, a benchmark for evaluating LM-mediated collaborative problem solving. Given a user request, PeopleJoin agents must identify teammates who might be able to assist, converse with these teammates to gather information, and finally compile a useful answer or summary for the original user. PeopleJoin comprises two evaluation domains: PeopleJoin-QA, focused on questions about tabular data, and PeopleJoin-DocCreation, focused on document creation tasks. The two domains are adapted from existing NLP benchmarks for database question answering and multi-document summarization; here, however, the information needed to complete these tasks is distributed across synthetic “organizations” of 2–20 users, simulating natural multi-user collaboration scenarios. We implemented several popular LM agent architectures, evaluating their accuracy and efficiency at completing tasks, and highlight new research questions that can be studied using PeopleJoin.
Large Language Models (LLMs) demonstrate exceptional proficiency in general-purpose tasks but struggle with numerical reasoning, particularly in low-resource languages like Bengali. Despite advancements, limited research has explored their numerical reasoning capabilities in these languages. To address this gap, we present BenNumEval (Bengali Numerical Evaluation), a benchmark designed to assess LLMs on numerical reasoning tasks in Bengali. It comprises six diverse tasks and a total of 3.2k samples curated from real-world problem-solving scenarios. Our extensive evaluations reveal that even with advanced prompting techniques such as Cross-Lingual Prompting (XLP) and Cross-Lingual Chain-of-Thought Prompting (XCoT), LLMs fall notably short of human-level performance, particularly when using Bengali Native Prompting (BNaP). These findings underscore the substantial gap between current LLM capabilities and human expertise in numerical reasoning, highlighting the need for more robust and linguistically inclusive AI models to advance Bengali Language Processing and equitable AI development. The source code for the system and evaluation pipeline is publicly available on GitHub.
Text-embedding models often exhibit biases arising from the data on which they are trained. In this paper, we examine a hitherto unexplored bias in text-embeddings: bias arising from the presence of \textit{names} such as persons, locations, organizations etc. in the text. Our study shows how the presence of \textit{name-bias} in text-embedding models can potentially lead to erroneous conclusions in the assessment of thematic similarity. Text-embeddings can mistakenly indicate similarity between texts based on names in the text, even when their actual semantic content has no similarity or indicate dissimilarity simply because of the names in the text even when the texts match semantically. We first demonstrate the presence of name bias in different text-embedding models and then propose \textit{text-anonymization} during inference which involves removing references to names, while preserving the core theme of the text. The efficacy of the anonymization approach is demonstrated on three downstream NLP tasks involving embedding similarities, achieving significant performance gains. Our simple and training-optimization-free approach offers a practical and easily implementable solution to mitigate name bias.
Recent years have witnessed rapid advancements in Large Language Models (LLMs). Nevertheless, it remains unclear whether state-of-the-art LLMs can infer the author of an anonymous research paper solely from the text, without any additional information. To investigate this novel challenge, which we define as Open-World Authorship Attribution, we introduce a benchmark comprising thousands of research papers across various fields to quantitatively assess model capabilities. Then, at the core of this paper, we tailor a two-stage framework to tackle this problem: candidate selection and authorship decision. Specifically, in the first stage, LLMs are prompted to generate multi-level key information, which are then used to identify potential candidates through Internet searches. In the second stage, we introduce key perspectives to guide LLMs in determining the most likely author from these candidates. Extensive experiments on our benchmark demonstrate the effectiveness of the proposed approach, achieving 60.7% and 44.3% accuracy in the two stages, respectively. We will release our benchmark and source codes to facilitate future research in this field.
Code-SPA: Style Preference Alignment to Large Language Models for Effective and Robust Code Debugging
PDF ↗Large language models (LLMs) have demonstrated impressive capabilities in coding tasks like code generation and debugging. However, code from real-world users is often poorly styled, containing various types of noise, such as structural inconsistencies, stylistic deviations and flawed test cases. To investigate this, we first simulate poorly styled code using eight types of code perturbations, and then demonstrate that the debugging performance of existing LLM-based methods significantly declines on such inputs. Furthermore, to address this, we propose a novel debugging method called Code-SPA, which aligns noisy code with the well-structured style familiar to LLMs, mitigating the impact of stylistic inconsistencies. Specifically, Code-SPA extracts the model’s preferred coding style from a reference snippet, then adjusts the input code by Concrete Syntax Tree (CST)-based transformations and LLM-assisted refinements before debugging. By aligning the code style preference, Code-SPA enhances the debugging performance of both code-specific and general-purpose LLMs on both poorly and well-styled code across the HumanEval, MBPP and EvalPlus datasets.
Tag-Instruct: Controlled Instruction Complexity Enhancement through Structure-based Augmentation
PDF ↗High-quality instruction data is crucial for developing large language models (LLMs), yet existing approaches struggle to effectively control instruction complexity. We present Tag-Instruct, a novel framework that enhances instruction complexity through structured semantic compression and controlled difficulty augmentation. Unlike previous prompt-based methods operating on raw text, Tag-Instruct compresses instructions into a compact tag space and systematically enhances complexity through RL-guided tag expansion. Through extensive experiments, we show that Tag-Instruct outperforms existing instruction complexity augmentation approaches. Our analysis reveals that operating in tag space provides superior controllability and stability across different instruction synthesis frameworks.
Language changes over time, including in the hate speech domain, which evolves quickly following social dynamics and cultural shifts. While NLP research has investigated the impact of language evolution on model training and has proposed several solutions for it, its impact on model benchmarking remains under-explored. Yet, hate speech benchmarks play a crucial role to ensure model safety. In this paper, we empirically evaluate the robustness of 20 language models across two evolving hate speech experiments, and we show the temporal misalignment between static and time-sensitive evaluations. Our findings call for time-sensitive linguistic benchmarks in order to correctly and reliably evaluate language models in the hate speech domain.
Look & Mark: Leveraging Radiologist Eye Fixations and Bounding boxes in Multimodal Large Language Models for Chest X-ray Report Generation
PDF ↗Recent advancements in multimodal Large Language Models (LLMs) have significantly enhanced the automation of medical image analysis, particularly in generating radiology reports from chest X-rays (CXR). However, these models still suffer from hallucinations and clinically significant errors, limiting their reliability in real-world applications. In this study, we propose Look & Mark (L&M), a novel grounding fixation strategy that integrates radiologist eye fixations (Look) and bounding box annotations (Mark) into the LLM prompting framework. Unlike conventional fine-tuning, L&M leverages in-context learning to achieve substantial performance gains without retraining. When evaluated across multiple domain-specific and general-purpose models, L&M demonstrates significant gains, including a 1.2% improvement in overall metrics (A.AVG) for CXR-LLaVA compared to baseline prompting and a remarkable 9.2% boost for LLaVA-Med. General-purpose models also benefit from L&M combined with in-context learning, with LLaVA-OV achieving an 87.3% clinical average performance (C.AVG)—the highest among all models, even surpassing those explicitly trained for CXR report generation. Expert evaluations further confirm that L&M reduces clinically significant errors (by 0.43 average errors per report), such as false predictions and omissions, enhancing both accuracy and reliability. These findings highlight L&M’s potential as a scalable and efficient solution for AI-assisted radiology, paving the way for improved diagnostic workflows in low-resource clinical settings.
Hyperbolic representations are effective in modeling knowledge graph data which is prevalently used to facilitate multi-hop reasoning. However, a rigorous and detailed comparison of the two spaces for this task is lacking. In this paper, through a simple integration of hyperbolic representations with an encoder-decoder model, we perform a controlled and comprehensive set of experiments to compare the capacity of hyperbolic space versus Euclidean space in multi-hop reasoning. Our results show that the former consistently outperforms the latter across a diverse set of datasets. In addition, through an ablation study, we show that a learnable curvature initialized with the delta hyperbolicity of the utilized data yields superior results to random initializations. Furthermore, our findings suggest that hyperbolic representations can be significantly more advantageous when the datasets exhibit a more hierarchical structure.