Wikipedia’s perceived high quality and broad language coverage have established it as a fundamental resource in NLP. However, in recent years, such assumptions of high quality have become the subject of scrutiny in low-resource and multilingual contexts. In this study, we subject the entirety of non-English Wikipedia to a data filtering procedure typically reserved for noisy web-text — a process which removes a large percentage of the collection’s data. In analysing the removed data, we reveal numerous systematic quality issues, such as script and language contamination, repeated template and placeholder articles, and a high concentration of bot-generated content. We consolidate these findings into a 4-level quality ranking of Wikipedia, which shows strong correspondence with alternative quality measures and heuristics. Lastly, we evaluate the downstream impact of quality filtering in three practical language modelling scenarios, showing that models trained on filtered data largely match or outperform those trained on raw Wikipedia, with the largest gains observed for lower-quality language editions. Ultimately, our experiments serve as a first step in establishing quality-aware best practices for Wikipedia utilization in NLP, laying groundwork that can inform future dataset creation and curation efforts.
论文检索
输入标题、作者或关键词，从 100,903 篇学术成果中精准定位
Retrieval-Augmented Generation (RAG) enhances the response capabilities of language models by integrating external knowledge sources. However, document chunking as an important part of RAG system often lacks effective evaluation tools. This paper first analyzes why existing RAG evaluation benchmarks are inadequate for assessing document chunking quality, specifically due to evidence sparsity. Based on this conclusion, we propose HiCBench, which includes manually annotated multi-level document chunking points, synthesized evidence-dense question answer(QA) pairs, and their corresponding evidence sources. We also propose HiChunk, a hierarchical document structuring framework using fine-tuned LLMs and the Auto-Merge retrieval algorithm to enhance retrieval quality. Experiments demonstrate that HiCBench effectively evaluates the impact of different chunking methods across the entire RAG pipeline. Moreover, HiChunk achieves better chunking quality within reasonable time consumption, thereby enhancing the overall performance of RAG systems. Source code is available at https://github.com/TencentCloudADP/hichunk.
The rapid advancement of large language models (LLMs) has driven the deployment of LLM-based AI tutors on online learning platforms. This widespread adoption highlights an urgent need for systematic benchmarks to evaluate their tutoring capabilities. However, existing evaluations predominantly focus on isolated, short-term interactions, overlooking the inherently long-term nature of learning. To bridge this gap, we introduce LongTutor, a benchmark for long-term personalized tutoring grounded in formative assessment theory. Built from expert-annotated real-world learning logs, LongTutor evaluates LLMs across three progressive tasks: historical evidence acquisition, knowledge state diagnosis, and adaptive teaching action. Our experiments reveal a critical capability mismatch: while LLMs excel at evidence acquisition, they struggle to effectively leverage long-term history for accurate diagnosis and adaptive teaching. To enable scalable benchmark expansion, we further propose an automated generator–verifier pipeline, paving the way toward truly long-term AI tutoring systems.
Supervised fine-tuning (SFT) on chain-of-thought (CoT) trajectories demonstrations is a common approach for enabling reasoning in large language models. Standard practices typically only retain trajectories with correct final answers (*positives*) while ignoring the rest (*negatives*). We argue that this paradigm discards substantial supervision and exacerbates overfitting, limiting out-of-domain (OOD) generalization. Specifically, we surprisingly find that incorporating *negative* trajectories into SFT yields substantial OOD generalization gains over *positive-only* training, as these trajectories often retain valid intermediate reasoning despite incorrect final answers. To understand this effect in depth, we systematically analyze data, training dynamics, and inference behavior, identifying 22 recurring patterns in negative chains that serve a dual role: they moderate loss descent to mitigate overfitting during training and boost policy entropy by 35.67% during inference to facilitate exploration. Motivated by these observations, we further propose **Gain-based LOss Weighting** (GLOW), an adaptive, sample-aware scheme that exploits such distinctive training dynamics by rescaling per-sample loss based on inter-epoch progress. Empirically, GLOW efficiently leverages unfiltered trajectories, yielding a 5.51% OOD gain over positive-only SFT on Qwen2.5-7B and boosting MMLU from 72.82% to 76.47% as an RL initialization. Code is available at [Github](https://github.com/Eureka-Maggie/GLOW).
Speculative decoding accelerates autoregressive generation by letting draft tokens bypass full verification, but conventional frameworks suffer from frequent false rejections, particularly when draft models produce semantically correct but lexically divergent outputs. In this paper, we present Calibrated Speculative Decoding (CSD), a training-free framework that recovers valid tokens discarded by standard verification. Guided by the principle of "Frequency-Guided Candidate Selection and Probability-Guarded Acceptance," CSD incorporates two lightweight modules: Online Correction Memory, which aggregates historical rejections to propose recurring divergence patterns as rescue candidates, and Semantic Consistency Gating, which verifies candidate admissibility using probability ratios instead of exact token matching. Our evaluation across diverse large language models demonstrates that CSD outperforms existing methods, achieving a peak throughput speedup of 2.33x. CSD preserves model accuracy across all tasks while further boosting performance on complex reasoning datasets. These results establish CSD as a highly effective, lightweight solution for practical LLM deployments.
Natural Language to Excel Formula (NL2Formula) translates user intent into executable spreadsheet formulas. However, current models often produce near-miss outputs—formulas that parse correctly yet fail at execution due to an incorrect function, operator, or reference. Through a systematic error analysis, we find that these errors repeatedly arise from a small set of structural decision points, motivating the need for typed error supervision rather than general error signals. To this end, we introduce an abstract syntax tree (AST)-based error taxonomy that organizes common error modes by the kind of decision that goes wrong in the parse tree. Building on this taxonomy, we propose Error-Aware Contrastive Few-Shot Learning (ECFL), an error-aware framework that unifies training and inference around typed error supervision. During offline training, ECFL mines near-miss errors, assigns error types under the taxonomy, and builds error-aware contrastive demonstrations for fine-tuning. During online inference, a lightweight predictor estimates likely error types and triggers targeted retrieval of contrastive demonstrations to guide single-pass decoding. Experiments show ECFL improves Exact Match (EM) by 6.4 points over supervised fine-tuning (SFT) and matches self-consistency (SC@5) accuracy at substantially lower inference cost.
Large language models (LLMs) have achieved strong performance in general machine translation, yet their ability in culture-aware scenarios remains poorly understood. To bridge this gap, we introduce CanMT, a Culture-Aware Novel-Driven Parallel Dataset for Machine Translation, together with a theoretically grounded, multi-dimensional evaluation frame work for assessing cultural translation quality. Leveraging CanMT, we systematically evaluate a wide range of LLMs and translation systems under different translation strategy constraints. Our findings reveal substantial performance disparities across models and demonstrate that translation strategies exert a systematic influence on model behavior. Further analysis shows that translation difficulty varies across types of culture-specific items, and that a persistent gap remains between models’recognition of culture-specific knowledge and their ability to correctly operationalize it in translation outputs. In addition, incorporating reference translations is shown to substantially improve evaluation reliability in LLM-as-a-judge, underscoring their essential role in assessing culture-aware translation quality. The corpus and code are available at CanMT.
Fine-tuning pretrained LLMs has proven effective for reaching state-of-the-art performance on specific tasks like machine translation. However, this process often implies sacrificing general-purpose capabilities, such as conversational reasoning and instruction-following, hampering the usefulness of the system in real-world applications requiring a mixture of skills. In this paper, we introduce Tower+, a suite of models designed to deliver strong performance on both translation and multilingual general-purpose text capabilities. We improve the TOwer (Alves et al., 2024) recipe by adding novel stages of preference optimization and reinforcement learning with verifiable rewards, in addition to continued pretraining and supervised fine-tuning. At each stage, we carefully generate and curate data to strengthen performance on translation and general-purpose tasks like coding, mathematics, and instruction-following. We develop models at multiple scales: 2B, 9B, and 72B. Our smaller models often outperform larger general-purpose open-weight and proprietary LLMs (e.g., Llama 3.3 70B, GPT-4o). Our largest model delivers best-in-class translation performance for high-resource languages, and top results on multilingual Arena Hard and IF-MT, a benchmark we introduce for evaluating both translation and instruction-following.
Medical multiple-choice question answering (MCQA) benchmarks implicitly assume that large language models (LLMs) should always commit to an answer. However, in clinical practice, uncertainty is pervasive and abstaining is often the safest action. We introduce MedQAbstain, a benchmark explicitly designed to evaluate medical abstention under uncertainty. MedQAbstain repurposes standard medical MCQA datasets by removing the gold answer and introducing an explicit "I abstain" option, framed as a safety-critical decision with clinical consequences. The benchmark supports systematic analysis across abstention regimes, distractor complexity, and input modalities, and elicits self-reported model confidence to study calibration. Across all settings, we find that state-of-the-art LLMs systematically overcommit, rarely abstaining even when the question itself is hidden. These results reveal a fundamental mismatch between LLM behavior and clinical norms, highlighting abstention as a critical but overlooked dimension of medical decision-making evaluation.
The semantic gap between colloquial user queries and professional legal documents presents a fundamental challenge in Legal Case Retrieval (LCR). Existing dense retrieval methods typically treat LCR as a black-box semantic matching process, neglecting the explicit juridical logic that underpins legal relevance. To address this, we propose GLIER (Generative Legal Inference and Evidence Ranking), a framework that reformulates retrieval as an inference process over latent legal variables. GLIER decomposes the task into two interpretability-driven stages: (1) A Joint Generative Inference module that translates raw queries into latent legal indicators (Charges and Legal Elements), employing a unified sequence-to-sequence strategy where charges and elements are generated jointly to enforce logical consistency; and (2) A Multi-View Evidence Fusion mechanism that aggregates generative confidence with structural and lexical signals for precise ranking. Extensive experiments on LeCaRD and LeCaRDv2 demonstrate that GLIER outperforms strong baselines like SAILER and KELLER. Notably, our framework exhibits exceptional data efficiency, maintaining robust performance even when trained with only 10% of the data.
DEFT: Demystifying VLN Failures via a Unified Dual-View Explainability Framework for LLM-based Agents
PDF ↗Large Language Models (LLMs) have emerged as central planners in Vision-and-Language Navigation (VLN), yet their complexity increasingly obscures their internal decision-making. Existing interpretability methods typically isolate temporal criticality from feature salience, creating an alignment gap and failing to account for the behavioral instability of black-box agents. To address this, we propose DEFT, a unified dual-view framework that demystifies agent behavior by jointly analyzing \textit{when} a decision is pivotal and \textit{what} visual evidence grounds it. Featuring a dual-head architecture with a shared latent representation, DEFT employs a \textit{Mask Head} for counterfactual-based criticality detection and an \textit{Action Head} that leverages an ensemble of surrogates to recover robust visual cues. Extensive experiments on MatterPort3D across three LLM-based agents demonstrate that DEFT outperforms baselines in both temporal and feature fidelity. User studies further validate its utility, showing 78% alignment with human intuition.
Modal dependency parsing-the task of identifying a semantic graph that represents who is responsible for an event-centered claim and with what degree of certainty-relies on recognizing source-introducing cues and correctly linking them to their associated content. However, prior work has largely focused on identifying sources only, treating cue expressions and their modal coverage as auxiliary signals. In this work, we propose a structured prediction framework that leverages large language models (LLMs) to explicitly identify source-cue pairs as well as their respective scope, which together define the modal contexts governing downstream source attribution for events. By concentrating learning at the source-cue level and constraining event-level decisions to a small, scope-defined candidate set, our top-down approach enables more efficient inference in long, event-rich documents. Experiments show this approach surpasses prior state-of-the-art results by 3 and 4% for English and Chinese datasets, respectively.
Temporal Evidence Chain for Temporal Knowledge Graph Question Answering with Large Language Models
PDF ↗Temporal Knowledge Graph Question Answering (TKGQA) aims to answer temporal questions using knowledge from Temporal Knowledge Graphs (TKGs).Existing LLM-based TKGQA methods typically utilize RAG-based or Agent-based paradigms, yet both struggle to construct reliable temporal evidence chains. RAG-based approaches primarily rely on semantic retrieval to fetch question-relevant contexts but overlook the structural dependencies within TKGs, leading to broken evidence chains, whereas iterative agents are prone to error propagation during multi-step reasoning.To address these limitations, we propose TECQA, a framework designed to construct temporal evidence chains for LLM reasoning. Firstly, TECQA employs structure-guided subgraph retrieval to capture structural dependencies and intermediate reasoning paths. Subsequently, it utilizes a k-nearest temporal neighbor pruning strategy to filter irrelevant noise while strictly preserving the continuous local history surrounding critical events. Finally, the retained temporal neighbors are serialized by temporal proximity to explicitly reconstruct a coherent temporal evidence chain. Extensive experiments on MultiTQ and CronQuestions demonstrate that TECQA achieves state-of-the-art performance, outperforming strong baselines by 45.3% particularly on complex queries. Code is available at https://github.com/SimonsLiu/TECQA.
Recently, activation steering has attracted considerable attention as a low-cost approach to improving the safety of large language models (LLMs). However, most existing methods apply interventions indiscriminately, often causing excessive refusal of benign queries. Although recent works have begun to explore selective intervention, their intervention decisions typically rely on residual stream activations where information is highly entangled, resulting in limited discriminative power and unreliable interventions. To address this issue, we propose FFN-Guided activation steering (FGAS). Motivated by the observation that feed-forward networks (FFNs) in LLMs serve as core modules for knowledge storage, we propose leveraging FFN output activations as more discriminative signals for intervention, since these activations more explicitly reflect the intent of a query. For a given query, FGAS projects the corresponding FFN output activation into a low-dimensional subspace that effectively separates harmful and benign queries, and then makes precise intervention decisions by assessing its similarity to pre-constructed prototype activations representing harmful and benign classes. Extensive experiments demonstrate that FGAS achieves state-of-the-art defense performance against various jailbreak attacks, while nearly preserving the model’s original performance on benign tasks.
Agentic Test-Time Scaling (TTS) has delivered state-of-the-art (SOTA) performance on complex software engineering tasks such as code generation and bug fixing. However, its practical adoption remains limited due to significant computational overhead, primarily driven by two key challenges: (1) the high cost associated with deploying excessively large ensembles, and (2) the lack of a reliable mechanism for selecting the optimal candidate solution—ultimately constraining the performance gains that can be realized. To address these challenges, we propose Entropy-Guided Stepwise Scaling (EGSS), a novel TTS framework that dynamically balances efficiency and effectiveness through entropy-guided adaptive search and robust test-suite augmentation.Extensive experiments on SWE-Bench-Verified demonstrate that EGSS consistently boosts performance by 5–10% across all evaluated models. Specifically, it increases the resolved ratio of Kimi-K2-Intruct from 63.2% to 72.2%, and GLM-4.6 from 65.8% to 74.6%. Furthermore, when paired with GLM-4.6, EGSS achieves new state-of-the-art among open-source large language models. In addition to these accuracy improvements, EGSS reduces inference-time token usage by over 28% compared to existing TTS methods, achieving simultaneous gains in both effectiveness and computational efficiency.
Training expert LLMs in domains with scarce fine-grained annotated data is admittedly challenging, often relying on multiple-choice questions (MCQs). However, standard outcome-based reinforcement learning (RL) on MCQs is risky. While outcome-based RL may improve accuracy, it frequently compromises the reasoning process, yielding internally inconsistent rationales that diverge from the final predictions. Existing solutions to supervise the reasoning process, such as large-scale Process Reward Models (PRMs), are prohibitively expensive. To address this, we propose CLARity, a cost-effective RL framework that enhances reasoning quality using a small, general-purpose LLM only. CLARity integrates a consistency-aware reward mechanism with a 2-stage refine-then-monitor training pipeline to enhance reasoning consistency, and a dynamic data reformulation strategy to better exploit annotated data available. Experiments demonstrate that CLARity can improve the consistency of responses by 16.5% over standard outcome-based RL, and bring an improvement of 7.5% in final accuracy. Human evaluations further confirm substantial gains in factual correctness and reasoning coherence, leading to more trustworthy model outputs. Thus, CLARity offers a generalizable solution that enables smaller models to effectively guide expert LLM training by monitoring reasoning consistency.
As large language models (LLMs) are more frequently used in retrieval-augmented generation pipelines, it is increasingly relevant to study their behavior under knowledge conflicts. Thus far, the role of the source of the retrieved information has gone unexamined. We address this gap with a novel framework to investigate how source preferences affect LLM resolution of inter-context knowledge conflicts in English, motivated by interdisciplinary research on credibility. By using synthetic sources, we study preferences for different types of sources without inheriting the biases of specific real-world sources. With a comprehensive, tightly-controlled evaluation of 13 open-weight LLMs, we find that LLMs prefer institutionally-corroborated information (e.g., government or newspaper sources) over information from people and social media. However, these source preferences can be reversed by simply repeating information from less credible sources. To mitigate repetition effects and maintain consistent preferences, we propose a novel method that reduces repetition bias by up to 79.2%, while also maintaining at least 72.5% of original preferences. We release all data and code to encourage future work on credibility and source preferences in knowledge-intensive NLP.
Zero-shot Jianzi Recognition as Structured Visual Information Extraction in Open Compositional Symbolic Systems
PDF ↗Guqin (古琴) Jianzi (減字) is an open and freely compositional tablature system that encodes performance actions rather than acoustic outcomes. Its automatic recognition remains largely unexplored, as conventional OCR assumes a closed and enumerable glyph set and struggles with Jianzi’s unbounded composition and manuscript-level variability.We introduce Zero-shot Jianzi Recognition, which formulates Jianzi recognition as vision-to-sequence prediction of canonical component sequences under a zero-shot split. To enable scalable supervision, we construct Synthetic-JZ from aligned online composition metadata. We then synthesize manuscript-like training images via component-wise style recomposition and manuscript-domain noise modeling, and fine-tune a vision–language model for end-to-end component sequence recognition. At inference time, a lightweight legality-guided correction module re-ranks decoding candidates, suppressing structural hallucinations without modifying the backbone.Experiments on two benchmarks show that our method achieves 63.02% sequence accuracy on Real-JZ, our manually annotated real-world Jianzi benchmark, surpassing Gemini-3-Pro by 35.11%. This result highlights the feasibility of reliable automated Jianzi recognition and its potential for large-scale digitization of historical Guqin Jianzi Pu manuscripts.
Mitigating Structural Knowledge Collapse in Domain-Specific LLMs via Morpheme-Aware KV-Aggregation
PDF ↗Standard tokenizers over-fragment domain terms, disrupting morpheme semantics. We characterize this representational misalignment as Structural Knowledge Collapse (SKC), where attention mechanisms fail to reconstruct coherent concepts from fragmented inputs. While existing input-centric solutions like vocabulary expansion address this, they necessitate expensive embedding retraining and neglect internal attention compositionality. To this end, we introduce Morpheme-aware KV-aggregation Attention (MorphKA), a lightweight adapter that dynamically consolidates fragments without tokenizer changes. Bypassing tokenizer retraining, MorphKA employs a dual-phase strategy, Input-Level Morpheme Aggregation (IMA) and Context-Aware KV-Aggregation (AMRF), to stabilize morpheme spans and synthesize higher-order concepts. Experiments on medical and legal benchmarks show MorphKA outperforms vocabulary adaptation baselines by 3.2–4.6%, reaching 7.9% on high-fragmentation terms. Moreover, MorphKA reduces catastrophic interference on general capabilities by 18–22% with ~80% fewer parameters than embedding retraining approaches.
SILO-BENCH: A Scalable Environment for Evaluating Distributed Coordination in Multi-Agent LLM Systems
PDF ↗Large language models are increasingly deployed in multi-agent systems to overcome context limitations by distributing information across agents. However, whether LLM-based agents can reliably coordinate when each observes only a fragment of the global problem remains unclear. Existing benchmarks often prescribe agent roles or interaction patterns, conflating coordination ability with role-based priors. We introduce SILO-BENCH, a role-free benchmark for evaluating free-form collaboration under information silos. The benchmark comprises 30 algorithmic tasks with exact ground-truth answers, organized into 3 complexity levels based on optimal communication complexity: aggregation, mesh, and global shuffle. To systematically probe coordination capabilities, we instantiate 54 configurations by varying 3 communication protocols, 6 agent scales and 3 frontier LLMs, conducting 1,620 experiments. We evaluate agent behavior along three dimensions: Success Rate, Token Consumption, and Communication Density. Our experiments reveal a fundamental Communication-Reasoning Gap: agents communicate actively, yet fail to translate interaction into effective distributed computation. Performance collapses as complexity increases, with Level-III tasks achieving zero success beyond 50 agents. These findings demonstrate that current LLMs cannot escape information silos through coordination alone. SILO-BENCH provides a foundation for tracking progress toward genuinely collaborative multi-agent systems. The code is available at https://github.com/jwyjohn/acl26-silo-bench.