This paper introduces GeoChain, a large-scale benchmark for evaluating step-by-step geographic reasoning in multimodal large language models (MLLMs). Leveraging 1.46 million Mapillary street-level images, GeoChain pairs each image with a 21-step chain-of-thought (CoT) question sequence (over 30 million Q&A pairs). These sequences guide models from coarse attributes to fine-grained localization across four reasoning categories - visual, spatial, cultural, and precise geolocation - annotated by difficulty. Images are also enriched with semantic segmentation (150 classes) and a visual locatability score. Our benchmarking of frontier MLLMs on a diverse 2,088-image subset reveals consistent challenges: models frequently exhibit weaknesses in visual grounding, display erratic reasoning, and struggle to achieve accurate localization, especially as the reasoning complexity escalates. GeoChain offers a robust diagnostic methodology, critical for fostering significant advancements in complex geographic reasoning within MLLMs.
论文检索
输入标题、作者或关键词，从 2,072 篇学术成果中精准定位
HomoGraphAdapter: A Homogeneous Graph Neural Network as an Effective Adapter for Vision-Language Models
PDF ↗Vision-Language Models (VLMs), such as CLIP, have exhibited significant advancements in recognizing visual concepts through natural language guidance. However, adapting these models to downstream tasks remains challenging. Existing adaptation methods either overlook the structural knowledge between the text and image modalities or create overly complex graphs containing redundant information for alignment, leading to suboptimal classification performance and increased computational overhead. This paper proposes a novel adapter-tuning methodology named Homogeneous Graph Adapter (HomoGraphAdapter), which transforms diverse textual and visual descriptions into a unified set of node representations and establishes edges between nodes for inter-modal and cross-modal semantic alignment. We leverage a straightforward homogeneous Graph Neural Network (GNN) to adapt positive and negative classifiers across text and image modalities. The classifiers comprehensively enhance the performance for few-shot classification and OOD generalization. Compared with the SOTA approach HeGraphAdapter, HomoGraphAdapter improves classification accuracy by an average of 1.51% for 1-shot and 0.74% for 16-shot on 11 datasets, while also reducing both precomputation time and training time.
Can Out-of-Distribution Evaluations Uncover Reliance on Prediction Shortcuts? A Case Study in Question Answering
PDF ↗A large body of recent work assesses models’ generalization capabilities through the lens of performance on out-of-distribution (OOD) datasets. Despite their practicality, such evaluations build upon a strong assumption: that OOD evaluations can capture and reflect upon possible failures in a real-world deployment. In this work, we challenge this assumption and confront the results obtained from OOD evaluations with a set of specific failure modes documented in existing question-answering (QA) models, referred to as a reliance on spurious features or prediction shortcuts.We find that different datasets used for OOD evaluations in QA provide an estimate of models’ robustness to shortcuts that have a vastly different quality, some largely under-performing even a simple, in-distribution evaluation. We partially attribute this to the observation that spurious shortcuts are shared across ID+OOD datasets, but also find cases where a dataset’s quality for training and evaluation is largely disconnected. Our work underlines limitations of commonly-used OOD-based evaluations of generalization, and provides methodology and recommendations for evaluating generalization within and beyond QA more robustly.
Refined Assessment for Translation Evaluation: Rethinking Machine Translation Evaluation in the Era of Human-Level Systems
PDF ↗As machine translation systems approach human-level quality, traditional evaluation methodologies struggle to detect subtle translation errors. We critically examine limitations in current gold-standard approaches (MQM and ESA), including inconsistencies from variable annotator expertise, excessive categorization complexity, coarse severity granularity, accuracy bias over fluency, and time constraints. To address this issue, we introduce a high-quality dataset consisting of human evaluations for English–Russian translations from WMT24, created by professional linguists. We show that expert assessments without time pressure yield substantially different results from standard evaluations. To enable consistent and rich annotation by these experts, we developed the RATE (Refined Assessment for Translation Evaluation) protocol. RATE provides a streamlined error taxonomy, expanded severity ratings, and multidimensional scoring balancing accuracy and fluency, facilitating deeper analysis of MT outputs. Our analysis, powered by this expert dataset, reveals that state-of-the-art MT systems may have surpassed human translations in accuracy while still lagging in fluency – a critical distinction obscured by existing accuracy-biased metrics. Our findings highlight that advancing MT evaluation requires not only better protocols but crucially, high-quality annotations from skilled linguists.
Beyond Single-User Dialogue: Assessing Multi-User Dialogue State Tracking Capabilities of Large Language Models
PDF ↗Large language models (LLMs) have demonstrated remarkable performance in zero-shot dialogue state tracking (DST), reducing the need for task-specific training. However, conventional DST benchmarks primarily focus on structured user-agent conversations, failing to capture the complexities of real-world multi-user interactions. In this study, we assess the robustness of LLMs in multi-user DST while minimizing dataset construction costs. Inspired by recent advances in LLM-based data annotation, we extend an existing DST dataset by generating utterances of a second user based on speech act theory. Our methodology systematically incorporates a second user’s utterances into conversations, enabling a controlled evaluation of LLMs in multi-user settings. Experimental results reveal a significant performance drop compared to single-user DST, highlighting the limitations of current LLMs in extracting and tracking dialogue states amidst multiple speakers. Our findings emphasize the need for future research to enhance LLMs for multi-user DST scenarios, paving the way for more realistic and robust DST models.
The prevailing “trivia-centered paradigm” for evaluating the cultural alignment of large language models (LLMs) is increasingly inadequate as these models become more advanced and widely deployed. Existing approaches typically reduce culture to static facts or values, testing models via multiple-choice or short-answer questions that treat culture as isolated trivia. Such methods neglect the pluralistic and interactive realities of culture, and overlook how cultural assumptions permeate even ostensibly “neutral” evaluation settings.In this position paper, we argue for intentionally cultural evaluation: an approach that systematically examines the cultural assumptions embedded in all aspects of evaluation, not just in explicitly cultural tasks. We systematically characterize the what, how, and circumstances by which culturally contingent considerations arise in evaluation, and emphasize the importance of researcher positionality for fostering inclusive, culturally aligned NLP research. Finally, we discuss implications and future directions for moving beyond current benchmarking practices, discovering important applications that we don’t know exist, and involving communities in evaluation design through HCI-inspired participatory methodologies.
LLM judges have gained popularity as an inexpensive and performant substitute for human evaluation. However, we observe that the meta-evaluation setting in which the reliability of these LLM evaluators is established is substantially different from their use in model development. To address this, we revisit meta-evaluations of LLM evaluators under a setting that more closely aligns with practice by examining evaluators’ ability to distinguish test system pairs that are closer in capability. Our fine-grained approach shows that all LLM evaluator’s correlations with human judgments are concerningly low when the models perform similarly, showcasing a key limitation of current norms. Equipped with this better methodology, we next analyze the impact that the choice of the reference model makes to LLM-as-a-judge evaluator performance. We show that single-reference evaluators only perform well at ranking test systems that fall within particular capability ranges, even if the standard meta-evaluation reports high overall correlation. Taken together, our analysis shows critical issues with current LLM meta-evaluation and recommend avenues for improvement.
We introduce MANTA, an automated pipeline that generates high-quality large-scale instruction fine-tuning datasets from massive web corpora while preserving their diversity and scalability. By extracting structured syllabi from web documents and leveraging high-performance LLMs, our approach enables highly effective query-response generation with minimal human intervention. Extensive experiments on 8B-scale LLMs demonstrate that fine-tuning on the MANTA-1M dataset significantly outperforms other massive dataset generation methodologies, particularly in knowledge-intensive tasks such as MMLU and MMLU-Pro, while also delivering superior performance across a broad spectrum of tasks. Moreover, MANTA supports seamless scalability by allowing the continuous integration of web corpus data, enabling expansion into domains requiring intensive knowledge.
MT-RewardTree: A Comprehensive Framework for Advancing LLM-Based Machine Translation via Reward Modeling
PDF ↗Process reward models (PRMs) have shown success in complex reasoning tasks for large language models (LLMs). However, their application to machine translation (MT) remains underexplored due to the lack of systematic methodologies and evaluation benchmarks. To address this gap, we introduce MT-RewardTree, a comprehensive framework for constructing, evaluating, and deploying process reward models in MT. Unlike traditional vanilla preference pair construction, we propose a novel method for automatically generating token-level preference pairs using approximate Monte Carlo Tree Search (MCTS), which mitigates the prohibitive cost of human annotation for fine-grained steps. Then, we establish the first MT-specific reward model benchmark and provide a systematic comparison of different reward modeling architectures, revealing that token-level supervision effectively captures fine-grained preferences. Experimental results demonstrate that our MT-PRM-Qwen-2.5-3B achieves state-of-the-art performance in both token-level and sequence-level evaluation given the same input prefix. Furthermore, we showcase practical applications where MT-PRMs successfully identify token-level translation differences and enable test-time alignment for LLMs without additional alignment training. Our work provides valuable insights into the role of reward models in MT research. Our code and data are released in https://sabijun.github.io/MT_RewardTreePage.
From n-gram to Attention: How Model Architectures Learn and Propagate Bias in Language Modeling
PDF ↗Current research on bias in language models (LMs) predominantly focuses on data quality, with significantly less attention paid to model architecture and temporal influences of data. Even more critically, few studies systematically investigate the origins of bias. We propose a methodology grounded in comparative behavioral theory to interpret the complex interaction between training data and model architecture in bias propagation during language modeling. Building on recent work that relates transformers to n-gram LMs, we evaluate how data, model design choices, and temporal dynamics affect bias propagation. Our findings reveal that: (1) n-gram LMs are highly sensitive to context window size in bias propagation, while transformers demonstrate architectural robustness; (2) the temporal provenance of training data significantly affects bias; and (3) different model architectures respond differentially to controlled bias injection, with certain biases (e.g. sexual orientation) being disproportionately amplified. As language models become ubiquitous, our findings highlight the need for a holistic approach- tracing bias to its origins across both data and model dimensions, not just symptoms, to mitigate harm.
Generation-Augmented Retrieval: Rethinking the Role of Large Language Models in Zero-Shot Relation Extraction
PDF ↗Recent advances in Relation Extraction (RE) emphasize Zero-Shot methodologies, aiming to recognize unseen relations between entities with no annotated data. Although Large Language Models (LLMs) have demonstrated outstanding performance in many NLP tasks, their performance in Zero-Shot RE (ZSRE) without entity type constraints still lags behind Small Language Models (SLMs). LLM-based ZSRE often involves manual interventions and significant computational overhead, especially when scaling to large-scale multi-choice data.To this end, we introduce RE-GAR-AD, which not only leverages the generative capability of LLMs but also utilizes their representational power without tuning LLMs. We redefine LLM-based ZSRE as a retrieval challenge, utilizing a Generation-Augmented Retrieval framework coupled with a retrieval Adjuster. Specifically, our approach guides LLMs through crafted prompts to distill sentence semantics and enrich relation labels. We encode sentences and relation labels using LLMs and match their embeddings in a triplet fashion. This retrieval technique significantly reduces token input requirements. Additionally, to further optimize embeddings, we propose a plug-in retrieval adjuster with only 2M parameters, which allows rapid fine-tuning without accessing LLMs’ parameters. Our LLM-based model demonstrates comparable performance on multiple benchmarks.
Large Language Models (LLMs) exhibit significant potential in complex software engineering tasks, however, their fault localization capabilities within repository are constrained by inherent limitations in max context length. Although Test-Time Scaling (TTS) can generate multiple candidate solutions, traditional selection strategies often fail to identify the optimal one. To solve this problem, we introduces Hierarchical Localization Reward Model (HiLoRM), which specifically designed to evaluate and select the most accurate fault localization candidates (at file, function, and line levels) from the multiple sampled outputs of LLMs, thereby enhancing localization accuracy. Furthermore, we constructed the HiFL-44k dataset, comprising approximately 44,000 fault localization instances, to train HiLoRM. Experimental results demonstrate that on the SWE-Bench-Lite dataset, HiLoRM improves the final line-level localization recall by 12% compared to a baseline model that does not use a reward model. Concurrently, HiLoRM exhibits a strong capability to evaluate predictions from larger LLMs (e.g., 32B parameters) and demonstrates transferability and generalization potential when applied to other fault localization methods. This work provides an effective methodology and an accessible model to significantly improve the accuracy and reliability of LLMs for repository-level fault localization. Our codes and datasets are available at https://github.com/SZU-ZJW/HiFL-Method.
Can Large Language Models (LLMs) simulate humans in making important decisions? Recent research has unveiled the potential of using LLMs to develop role-playing language agents (RPLAs), mimicking mainly the knowledge and tones of various characters. However, imitative decision-making necessitates a more nuanced understanding of personas. In this paper, we benchmark the ability of LLMs in persona-driven decision-making. Specifically, we investigate whether LLMs can predict characters’ decisions provided by the preceding stories in high-quality novels. Leveraging character analyses written by literary experts, we construct a dataset LIFECHOICE comprising 2,512 characters’ decision points from 470 books. Then, we conduct comprehensive experiments on LIFECHOICE with various LLMs and RPLA methodologies. The results demonstrate that state-of-the-art LLMs exhibit promising capabilities in this task, yet substantial room for improvement remains. Hence, we further propose the CHARMAP method, which adopts persona-based memory retrieval and significantly advances RPLAs on this task.
Methods for story generation with Large Language Models (LLMs) have come into the spotlight recently. We create a novel taxonomy of LLMs for story generation consisting of two major paradigms: (i) independent story generation by an LLM, and (ii) author-assistance for story generation – a collaborative approach with LLMs supporting human authors. We compare existing works based on their methodology, datasets, generated story types, evaluation methods, and LLM usage. With a comprehensive survey, we identify potential directions for future work
A conflict of interest (COI) appears when a person or a company has two or more interests that may directly conflict. This happens, for instance, when a scientist whose research is funded by a company audits the same company. For transparency and to avoid undue influence, public repositories of relations of interest are increasingly recommended or mandated in various domains, and can be used to avoid COIs. In this work, we propose an LLM-based open information extraction (OpenIE) framework for extracting financial or other types of interesting relations from scientific text. We target scientific publications in which authors declare funding sources or collaborations in the acknowledgment section, in the metadata, or in the publication, following editors’ requirements. We introduce an extraction methodology and present a knowledge base (KB) with a comprehensive taxonomy of COI centric relations. Finally, we perform a comparative study of disclosures of two journals in the field of toxicology and pharmacology.
SampleMix: A Sample-wise Pre-training Data Mixing Strategy by Coordinating Data Quality and Diversity
PDF ↗Existing pretraining data mixing methods for large language models (LLMs) typically follow a domain-wise methodology, a top-down process that first determines domain weights and then performs uniform data sampling across each domain. However, these approaches neglect significant inter-domain overlaps and commonalities, failing to control the global diversity of the constructed training dataset. Further, uniform sampling within domains ignores fine-grained sample-specific features, potentially leading to suboptimal data distribution. To address these shortcomings, we propose a novel sample-wise data mixture approach based on a bottom-up paradigm. This method performs global cross-domain sampling by systematically evaluating the quality and diversity of each sample, thereby dynamically determining the optimal domain distribution. Comprehensive experiments across multiple downstream tasks and perplexity assessments demonstrate that SampleMix surpasses existing domain-based methods. Meanwhile, SampleMix requires 1.4x to 2.1x fewer training steps to achieve the baselines’ performance, highlighting the substantial potential of SampleMix to optimize pre-training data.
Active Domain Knowledge Acquisition with 100-Dollar Budget: Enhancing LLMs via Cost-Efficient, Expert-Involved Interaction in Sensitive Domains
PDF ↗Large Language Models (LLMs) have demonstrated an impressive level of general knowledge. However, they often struggle in highly specialized and sensitive domains such as drug discovery and rare disease research due to the lack of expert knowledge, which is often costly to obtain. In this paper, we propose a novel framework (PU-ADKA) designed to efficiently enhance domain-specific LLMs by actively engaging domain experts within a fixed budget. Unlike traditional fine-tuning approaches, PU-ADKA proactively identifies and queries the most appropriate expert from a team, taking into account each expert’s availability, competency, knowledge boundaries, and consultation cost. We train PU-ADKA using simulations on PubMed publication data and validate it through domain expert interactions, showing promising improvements in LLM domain knowledge acquisition. Furthermore, our experiments with a real-world drug development team validate that PU-ADKA can significantly enhance LLM performance in specialized domains while adhering to strict budget constraints. In addition to outlining our methodological innovations and experimental results, we release a new benchmark dataset, CKAD, for cost-effective LLM domain knowledge acquisition to foster further research in this challenging area.
DaMoC: Efficiently Selecting the Optimal Large Language Model for Fine-tuning Domain Tasks Based on Data and Model Compression
PDF ↗Large language models (LLMs) excel in general tasks but struggle with domain-specific ones, requiring fine-tuning with specific data. With many open-source LLMs available, selecting the best model for fine-tuning downstream tasks is challenging, primarily focusing on how to quickly identify the optimal LLM. We introduce a Data and Model Compression Framework (DaMoC) that addresses this challenge by: 1) Data Level: A systematic categorization of data filtering methodologies for LLMs is first established, classifying them into three distinct paradigms: (1) distribution-aware methods, (2) quality-aware methods, and (3) hybrid approaches considering both dimensions. Further, we enhance the density of key tokens in the text achieving token compression. Subsequently, we use an LLM to iterative rewrite the text to optimize its expression. 2) Model Level: We use layer similarity scores to assess each layer’s importance and remove those with lower importance. Then, we introduce a sparse merging paradigm to preserve as much of the original model’s capability as possible. Extensive experiments on four datasets, medical Q&A, financial Q&A, general Q&A, and reading comprehension, show that we can select the optimal LLM while saving approximately 20-fold in training time.
Large Language Models (LLMs) have demonstrated significant potential in decision-making and reasoning, particularly when integrated with various tools to effectively solve complex problems. However, existing benchmarks for evaluating LLMs’ tool usage face several limitations: (1) limited evaluation scenarios, often lacking assessments in real multi-turn dialogue contexts; (2) narrow evaluation dimensions, with insufficient detailed assessments of how LLMs use tools; and (3) reliance on LLMs or real API executions for evaluation, which introduces significant overhead. To address these challenges, we introduce ACEBench, a comprehensive benchmark for assessing tool usage in LLMs. ACEBench categorizes data into three primary types based on evaluation methodology: Normal, Special, and Agent. “Normal” evaluates tool usage in basic scenarios; “Special” evaluates tool usage in situations with ambiguous or incomplete instructions; “Agent” evaluates tool usage through multi-agent interactions to simulate real-world, multi-turn dialogues. We conducted extensive experiments using ACEBench, analyzing various LLMs in-depth and providing a more granular examination of error causes across different data types.
DPF-CM: A Data Processing Framework with Privacy-Preserving Vector Databases for Chinese Medical LLMs Training and Deployment
PDF ↗Current open-source training pipelines for Chinese medical language models predominantly emphasize optimizing training methodologies to enhance the performance of large language models (LLMs), yet lack comprehensive exploration into training data processing. To address this gap, we propose DPF-CM, a holistic Data Processing Framework for Chinese Medical LLMs training and deployment. DPF-CM comprises two core modules. The first module is a data processing pipeline tailored for model training. Beyond standard data processing operations, we (1) introduce a chained examples context-learning strategy to generate question-oriented instructions to mitigate the lack of instruction content, and (2) implement an ensemble-based filtering mechanism for preference data curation that averages multiple reward models to suppress noisy samples. The second module focuses on privacy preservation during model deployment. To prevent privacy risks from the inadvertent exposure of training data, we propose a Privacy Preserving Vector Database (PPVD) approach, which involves model memory search, high-risk database construction, secure database construction, and match-and-replace, four key stages to minimize privacy leakage during inference collectively. Experimental results show that DPF-CM significantly improves model accuracy, enabling our trained Chinese medical LLM to achieve state-of-the-art performance among open-source counterparts. Moreover, the framework reduces training data privacy leakage by 27%.