Searcharxiv⌕ Search

arXiv subjects

Jingwei Ni

Publications and source records attributed to Jingwei Ni.

At least 19 recordsLinked to original sources

Same Beliefs, Different Policies: Learning Reusable Representations for Early Stopping in LLM Reasoning

Reasoning LLMs often continue reasoning after reaching a correct answer, wasting computation and sometimes replacing it with an incorrect one. Deciding when to stop is a metareasoning problem: the model must weigh what further reasoning could achieve against its cost, given what it can already answer. Direct utility training learns a stopping representation at a chosen token price; supporting new prices can require additional controller training. We introduce BELT (BELief-based Termination), which separates outcome prediction from stopping control. It learns beliefs about current and future answers from a frozen LLM's hidden states, then fits a linear stopping policy on the frozen features. The prediction targets are independent of token price, so one probe supports different accuracy-cost trade-offs through CPU policy refitting or threshold adjustment. Averaged over five out-of-distribution (OOD) math benchmarks, BELT reduces generated tokens by 41% on Qwen3-1.7B and 51% on OLMo-3-7B-Think without loss of accuracy. The same math-trained BELT transfers to coding and agentic tasks without training on code or agentic data. In all three domains, its accuracy-cost trade-offs are competitive with or better than those of end-to-end controllers. On OLMo3, its probe has 21.2x fewer parameters than one checkpoint of the strongest baseline, OS-Pruner, and needs 55x fewer training GPU-hours than OS-Pruner's five price-specific checkpoints.

cs.AI↗

Climate-ModernBERT: Revisiting Corpus Composition for Domain-Adaptive Continued Pretraining

Natural Language Processing (NLP) in the climate domain requires models to process heterogeneous text sources, including scientific literature, policy disclosures, and synthetic reports. However, how to effectively combine diverse domain corpora during continued pretraining (CPT) remains underexplored. We introduce Climate-ModernBERT, a family of climate-adapted encoder models obtained through continued pretraining of ModernBERT-Base on three climate corpora: academic climate text, climate-filtered web data, and synthetic climate documents. We systematically compare joint continued pretraining on corpus mixtures with parameter-space merging of independently specialized checkpoints. Across nine climate NLP benchmarks, our best model achieves 76.3 average F_1, improving significantly over a vanilla ModernBERT baseline by 2.8 points. Within the climate NLP setting, the results show that academic climate corpora provide the strongest adaptation signal among the evaluated sources, while parameter-space merging improves over joint multi-source training and better preserves complementary information from heterogeneous climate corpora. We release all Climate-ModernBERT variants and training checkpoints to support future research in climate NLP and domain-adaptive pretraining.

cs.CL↗

Understanding Failures in LLM Reasoning by Learning Structured Representations of Chain-of-Thought

Chain-of-Thought (CoT) is central to improving reasoning in large language models, yet existing evaluation methods focus on step-level signals and ignore the underlying structure of reasoning traces. This raises a key question: what aspects of reasoning quality (i.e., correctness, efficiency, and failure patterns) are encoded in the structure of a CoT trace? We introduce \textbf{CoT2Graph}, a framework that represents CoT traces as directed dependency graphs, making relationships between reasoning steps explicit and capturing patterns such as backtracking, branching, and subtask merging. We model these graphs with graph neural networks and use attribution-based explanations to identify structural patterns associated with reasoning quality. CoT2Graph reveals distinct patterns of successful and failed reasoning: correct CoTs place greater importance on early planning, whereas incorrect CoTs are associated with overloaded synthesis steps and dependency violations. We further show that graph-based signals better predict whether a CoT leads to a correct answer, outperforming previous tree-based structural methods by up to 2.5\% and simple length-based baselines by 7.8\%, while also improving Best-of-$N$ selection. Overall, our results demonstrate that reasoning structure encodes rich signals about reasoning quality, motivating a shift from sequence-level evaluation to structural analysis.

cs.CL↗

When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturation

Artificial intelligence benchmarks are an important mechanism to measure model progress and guide deployment decisions. However, benchmarks quickly "saturate", making it difficult to differentiate models and diminishing their long-term value. In this study, we define benchmark saturation and analyze it across 60 language model benchmarks using 14 properties that relate to saturation. We find that nearly half of our benchmarks exhibit saturation, with rates increasing with age. Further, we find that resilience to saturation is impacted by expert-curation, not by public test data. Our results suggest that design choices can extend benchmark longevity and inform more durable evaluation approaches.

cs.AI↗

Skill Self-Play: Pushing the Frontier of LLM Capability with Co-Evolving Skills

LLM training is shifting from manual design and annotation to interaction-driven self-evolution. However, existing self-evolutionary methods face a fundamental dilemma between task diversity and verification reliability: environment-bound methods obtain precise feedback but confine learning to narrow domains, while open-ended self-generation broadens the task space but lacks reliable verification, allowing misleading rewards to pollute the training loop. We identify agent skills as a powerful middle ground to reconcile this tension: each skill ensures deep, verifiable execution in a specific scenario, while dynamic routing across skills maintains open-ended task variety. Leveraging this insight, we introduce Skill Self-Play (Skill-SP), a co-evolutionary framework comprising a proposer, a solver, and a dynamic skill controller. Orchestrated via a reinforcement learning loop, these components co-evolve in a continuous self-play loop: the proposer generates challenging tasks conditioned on dynamically sampled skills; the solver explores candidate solutions to push its capability boundaries; and the skill controller collects execution feedback to update and expand the skill library. This interactive co-evolution effectively bridges the gap between structured verification and open-ended exploration. Empirical evaluations on tool-use and reasoning benchmarks demonstrate that Skill-SP, serving as a robust evolution engine, consistently pushes the performance ceiling of competent backbones while catalyzing striking turnarounds for initially misaligned models. Our code is available at https://github.com/Qwen-Applications/skill-self-play.

cs.CL↗

REKEY: Metadata-Grounded Visual-Key Regeneration for Contamination-Resilient VQA Evaluation

Static visual question answering (VQA) benchmarks age quickly: Once the items leak into training corpora, scores can reflect memorization rather than genuine visual ability, thus obscuring real progress. Rebuilding high-quality benchmarks such as V*Bench requires substantial human annotation, yet each static release can quickly become another leaked artifact. We propose ReKey, a live benchmark protocol that randomly regenerates the answer-bearing local detail, or visual key, in real images at evaluation time. Using human-validated edit slots, ReKey samples fresh instances with new answers, construction-grounded labels, and controlled visual-search difficulty. On V*Bench, the ReKey regenerated benchmark reveals a sharp score jump across eight frontier vision-language models (VLMs): The original items score 9.5--18.8 percentage points higher than the regenerated variants. By making the visual key renewable, ReKey keeps evaluation fresh as models and training data evolve.

cs.CV↗

GD$^2$PO: Mitigating Multi-Reward Conflicts via Group-Dynamic reward-Decoupled Policy Optimization

As LLMs advance, post-training reinforcement learning (RL) increasingly relies on multi-dimensional rewards to cultivate comprehensive capabilities. This shift demands new algorithms capable of optimizing diverse and potentially competing objectives simultaneously. To address this, existing methods such as Group reward-Decoupled Policy Optimization (GDPO) decompose the overall score into independent reward groups, then compute the RL loss separately within each group. However, this strategy still encounters multi-reward conflicts: a single rollout can yield positive advantages on certain reward dimensions but negative ones on others, causing opposing signals to cancel each other out during aggregation, further hindering RL training efficiency. Inspired by Dynamic sAmpling Policy Optimization (DAPO), which improves RL training efficiency by filtering out ineffective rollouts with near-zero advantages, we propose Group-Dynamic reward-Decoupled Policy Optimization (GD$^2$PO). Specifically, GD$^2$PO employs a conflict-aware filtering mechanism to mask out rollouts suffering from severe reward-wise disagreement. By preventing conflicting signals from canceling each other out, this masking strategy preserves and enhances the magnitude of effective RL advantages, thereby significantly accelerating learning efficiency. Furthermore, we introduce query-level reweighting to dynamically adjust the update intensity of each query based on its overall reward consensus. Experiments on various multi-reward scenarios, including tool calling and human preference alignment, demonstrate that GD$^2$PO consistently and significantly outperforms existing baselines. The code is available at https://github.com/Qwen-Applications/GD2PO.

cs.LG↗

ThinkBooster: A Unified Framework for Seamless Test-Time Scaling of LLM Reasoning

Test-time compute (TTC) scaling has emerged as a powerful paradigm for improving large language model (LLM) reasoning by allocating additional compute during inference, e.g., via multi-sample generation and verifier-based reranking. Existing TTC scaling strategies and reasoning scorers remain fragmented, evaluated under inconsistent protocols, and are rarely analyzed through the lens of quality-cost trade-offs. We introduce ThinkBooster, a unified framework for seamless test-time compute scaling of LLM reasoning, which consists of (i) a modular Python library implementing state-of-the-art TTC scaling strategy and scorer families, (ii) a benchmark that jointly evaluates performance and computational efficiency, and (iii) a deployable OpenAI-compatible proxy service that enables drop-in integration of adaptive reasoning into real-world applications. We further provide a demo visual debugger for inspecting the reasoning trajectories, intermediate selection decisions, and alternative reasoning paths. Empirical results on mathematical and coding tasks reveal the performance-compute trade-offs of TTC scaling strategies and scoring methods and demonstrate that ThinkBooster provides practical gains in real-world tasks. The code is available online under an MIT license.

cs.CL↗

Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skills

Large Language Model (LLM) agents increasingly rely on domain-specific skills, yet manually authoring such skills does not scale, and skills generated purely from parametric knowledge often miss critical operational pitfalls. We introduce Trace2Skill, a framework that consolidates broad execution trajectories in parallel into a unified skill directory through inductive reasoning over agent experience. Trace2Skill supports both deepening existing human-written skills and creating useful skills from weak LLM-generated drafts. Experiments demonstrate the effectiveness of Trace2Skill across diverse domains, including office workflows, math reasoning, and vision QA. Importantly, the evolved skills are not merely memorized artifacts of the trajectories used to create them: they often transfer across model scales, across model families, and to out-of-distribution settings. For example, skills evolved from Qwen3.5-35B trajectories improve a Qwen3.5-122B agent by up to $57.65$ percentage points on WikiTableQuestions. Further analyses show that Trace2Skill outperforms sequential skill editing and ReasoningBank-style retrieval memories, compresses recurring failures and workarounds into standard operating procedures (SoPs), and yields portable skills that can be reused without parameter updates or test-time retrieval.

cs.AI↗

Skill-RM: Unifying Heterogeneous Evaluation Criteria via Agent Skill

Reward models (RMs) provide critical feedback signals for LLM post-training, notably in reinforced fine-tuning (RFT) and reinforcement learning (RL) pipelines. However, current reward evaluation relies on heterogeneous criteria such as rule-based verifiers, ground-truth references, procedural checklists, and complex rubrics, where a unified mechanism to integrate all types of evidence remains unexplored. To this end, we propose Skill Reward Model (Skill-RM), a unified framework that reformulates reward modeling as the execution of a reusable Reward-Evaluation Skill. By treating reward computation as a structured agentic task, Skill-RM provides a consistent interface to orchestrate heterogeneous resources, dynamically selecting and aggregating evidence tailored to the specific requirements of each input. This approach enables the reward model to move beyond static evaluation, ensuring consistency and transparency across diverse tasks. Extensive experiments on reward benchmarks and downstream applications, including best-of-N selection and reinforcement learning, demonstrate that Skill-RM consistently outperforms traditional judge baselines. Our findings suggest that Skill-RM not only provides a unified solution for reward modeling but also achieves superior performance through the strategic and dynamic orchestration of evidence. The code is at https://github.com/Qwen-Applications/Skill-RM.

cs.LG↗

Tackling the Root of Misinformation by Teaching Laypeople about Logical Fallacies via Socratic Questioning and Critical Argumentation

Identifying logical fallacies in everyday discourse is challenging for many people. This challenge is amplified in the era of Large Language Models (LLMs), where malicious agents can deploy fallacious arguments to disseminate misinformation at scale. In this work, we explore the potential of LLMs as part of the solution. We introduce LFTutor, an intelligent tutoring system which uses LLMs to tutor laypeople and help them learn about logical fallacies. LFTutor integrates intent-driven Socratic questioning and critical argumentation principles to actively engage learners to reflect on their reasoning. Through both automatic and human evaluations, we demonstrate that LFTutor significantly outperforms baseline LLMs lacking these pedagogical strategies. This work highlights the promise of combining LLMs with pedagogical scaffolding to foster critical thinking and argument literacy in the age of AI.

cs.AI↗

ReProbe: Efficient Test-Time Scaling of Multi-Step Reasoning by Probing Internal States of Large Language Models

LLMs can solve complex tasks by generating long, multi-step reasoning chains. Test-time scaling (TTS) can further improve performance by sampling multiple variants of intermediate reasoning steps, verifying their correctness, and selecting the best steps for continuation. However, existing verification approaches, such as Process Reward Models (PRMs), are computationally expensive and require large-scale human or model-generated annotations. We propose a lightweight alternative for step-level reasoning verification based on probing the internal states of LLMs. We train a transformer-based probe that uses the internal states of a frozen LLM to estimate the credibility of its reasoning steps during generation. Annotation can be provided either by a larger LLM (e.g., DeepSeek-R1) or in a self-supervised manner by the original model itself. The probes are lightweight, containing fewer than 10M parameters. Across multiple domains, including mathematics, planning, and general knowledge question answering, our probes match or exceed the performance of PRMs that are up to 810x larger. These results suggest that LLM internal states encode confidence in their reasoning processes and can serve as reliable signals for step verification, offering a promising path toward scalable, generalizable TTS and more introspective LLMs.

cs.AI↗

WeatherArchive-Bench: Benchmarking Retrieval-Augmented Reasoning for Historical Weather Archives

Historical archives on weather events are collections of enduring primary source records that offer rich, untapped narratives of how societies have experienced and responded to extreme weather events. These qualitative accounts provide insights into societal vulnerability and resilience that are largely absent from meteorological records, making them valuable for climate scientists to understand societal responses. However, their vast scale, noisy digitized quality, and archaic language make it difficult to transform them into structured knowledge for climate research. To address this challenge, we introduce WeatherArchive-Bench, the first benchmark for evaluating retrieval-augmented generation (RAG) systems on historical weather archives. WeatherArchive-Bench comprises two tasks: WeatherArchive-Retrieval, which measures a system's ability to locate historically relevant passages from over one million archival news segments, and WeatherArchive-Assessment, which evaluates whether Large Language Models (LLMs) can classify societal vulnerability and resilience indicators from extreme weather narratives. Extensive experiments across sparse, dense, and re-ranking retrievers, as well as a diverse set of LLMs, reveal that dense retrievers often fail on historical terminology, while LLMs frequently misinterpret vulnerability and resilience concepts. These findings highlight key limitations in reasoning about complex societal indicators and provide insights for designing more robust climate-focused RAG systems from archival contexts. The constructed dataset and evaluation framework are publicly available at https://anonymous.4open.science/r/WeatherArchive-Bench/.

cs.CL↗

LEXam: Benchmarking Legal Reasoning on 340 Law Exams

Long-form legal reasoning remains a key challenge for large language models (LLMs) in spite of recent advances in test-time scaling. To address this, we introduce LEXam, a novel benchmark derived from 340 law exams spanning 116 law school courses across a range of subjects and degree levels. The dataset comprises 7,537 law exam questions in English and German. It includes both long-form, open-ended questions and multiple-choice questions with varying numbers of options. Besides reference answers, the open questions are also accompanied by explicit guidance outlining the expected legal reasoning approach such as issue spotting, rule recall, or rule application. Our evaluation on both open-ended and multiple-choice questions present significant challenges for current LLMs; in particular, they notably struggle with open questions that require structured, multi-step legal reasoning. Moreover, our results underscore the effectiveness of the dataset in differentiating between models with varying capabilities. Deploying an ensemble LLM-as-a-Judge paradigm with rigorous human expert validation, we demonstrate how model-generated reasoning steps can be evaluated consistently and accurately, closely aligning with human expert assessments. Our evaluation setup provides a scalable method to assess legal reasoning quality beyond simple accuracy metrics. Project page: https://lexam-benchmark.github.io/.

cs.CL↗

Can Reasoning Help Large Language Models Capture Human Annotator Disagreement?

Variation in human annotation (i.e., disagreements) is common in NLP, often reflecting important information like task subjectivity and sample ambiguity. Modeling this variation is important for applications that are sensitive to such information. Although RLVR-style reasoning (Reinforcement Learning with Verifiable Rewards) has improved Large Language Model (LLM) performance on many tasks, it remains unclear whether such reasoning enables LLMs to capture informative variation in human annotation. In this work, we evaluate the influence of different reasoning settings on LLM disagreement modeling. We systematically evaluate each reasoning setting across model sizes, distribution expression methods, and steering methods, resulting in 60 experimental setups across 3 tasks. Surprisingly, our results show that RLVR-style reasoning degrades performance in disagreement modeling, while naive Chain-of-Thought (CoT) reasoning improves the performance of RLHF LLMs (RL from human feedback). These findings underscore the potential risk of replacing human annotators with reasoning LLMs, especially when disagreements are important.

cs.CL↗

pdfQA: Diverse, Challenging, and Realistic Question Answering over PDFs

PDFs are the second-most used document type on the internet (after HTML). Yet, existing QA datasets commonly start from text sources or only address specific domains. In this paper, we present pdfQA, a multi-domain 2K human-annotated (real-pdfQA) and 2K synthetic dataset (syn-pdfQA) differentiating QA pairs in ten complexity dimensions (e.g., file type, source modality, source position, answer type). We apply and evaluate quality and difficulty filters on both datasets, obtaining valid and challenging QA pairs. We answer the questions with open-source LLMs, revealing existing challenges that correlate with our complexity dimensions. pdfQA presents a basis for end-to-end QA pipeline evaluation, testing diverse skill sets and local optimizations (e.g., in information retrieval or parsing).

cs.CL↗

Apertus: Democratizing Open and Compliant LLMs for Global Language Environments

We present Apertus, a fully open suite of large language models (LLMs) designed to address two systemic shortcomings in today's open model ecosystem: data compliance and multilingual representation. Unlike many prior models that release weights without reproducible data pipelines or regard for content-owner rights, Apertus models are pretrained exclusively on openly available data, retroactively respecting `robots.txt` exclusions and filtering for non-permissive, toxic, and personally identifiable content. To mitigate risks of memorization, we adopt the Goldfish objective during pretraining, strongly suppressing verbatim recall of data while retaining downstream task performance. The Apertus models also expand multilingual coverage, training on 15T tokens from over 1800 languages, with ~40% of pretraining data allocated to non-English content. Released at 8B and 70B scales, Apertus approaches state-of-the-art results among fully open models on multilingual benchmarks, rivalling or surpassing open-weight counterparts. Beyond model weights, we release all scientific artifacts from our development cycle with a permissive license, including data preparation scripts, checkpoints, evaluation suites, and training code, enabling transparent audit and extension.

cs.CL↗

Co-DETECT: Collaborative Discovery of Edge Cases in Text Classification

We introduce Co-DETECT (Collaborative Discovery of Edge cases in TExt ClassificaTion), a novel mixed-initiative annotation framework that integrates human expertise with automatic annotation guided by large language models (LLMs). Co-DETECT starts with an initial, sketch-level codebook and dataset provided by a domain expert, then leverages the LLM to annotate the data and identify edge cases that are not well described by the initial codebook. Specifically, Co-DETECT flags challenging examples, induces high-level, generalizable descriptions of edge cases, and assists user in incorporating edge case handling rules to improve the codebook. This iterative process enables more effective handling of nuanced phenomena through compact, generalizable annotation rules. Extensive user study, qualitative and quantitative analyses prove the effectiveness of Co-DETECT.

cs.CL↗