As AI coding agents take on increasingly complex, long-horizon software engineering tasks, existing benchmarks are rapidly saturating and their evaluation quality has come under serious scrutiny: a recent audit found that nearly 60% of unsolved SWE-bench Verified instances contain flawed tests -- either overly narrow tests that reject correct solutions or overly broad tests that check unstated requirements -- and that frontier models can verbatim reproduce gold patches from training data. Code refactoring, which requires coordinated, behavior-preserving changes across many files, offers a substantially harder and more realistic test of agent capability, yet remains underserved by current benchmarks. We introduce SWE-Bench ProMax, an expert-curated, multilingual code refactoring benchmark of 170 instances drawn from real commits across seven programming languages (Python, Java, TypeScript, Go, C, C++, and Rust). Every instance undergoes rigorous, multi-stage curation that directly addresses the quality problems identified in prior benchmarks: issue descriptions are rewritten from scratch to provide precise, unambiguous specifications, and test suites are manually reviewed to remove overly narrow and overly broad tests. Tasks with insufficient complexity or limited cross-file scope are filtered out, yielding a benchmark of challenging, large-scale refactoring tasks that average 11.5 modified files and 261.6 lines of code per instance, substantially exceeding the scale of existing benchmarks. Experiments with frontier models under two agent scaffolds show that the best model achieves only 41.2% resolve rate, confirming that SWE-Bench ProMax presents a meaningful and unsaturated challenge for current AI coding agents. Our benchmark is available at https://huggingface.co/datasets/swe-bench-promax/SWE-Bench-ProMax.
Multimodal Large Language Models (MLLMs) have achieved remarkable performance in Visually Rich Document Understanding (VRDU) tasks, but their capabilities are mainly evaluated on pristine, well-structured document images. We consider content restoration from shredded fragments, a challenging VRDU setting that requires integrating visual pattern recognition with semantic reasoning under significant content discontinuities. To facilitate systematic evaluation of complex VRDU tasks, we introduce ShredBench, a benchmark supported by an automated generation pipeline that renders fragmented documents directly from Markdown. The proposed pipeline ensures evaluation validity by allowing the flexible integration of latest or unseen textual sources to prevent training data contamination. ShredBench assesses four scenarios (English, Chinese, Code, Table) with three fragmentation granularities (8, 12, 16 pieces). Empirical evaluations on state-of-the-art MLLMs reveal a significant performance gap: The method is effective on intact documents; however, once the document is shredded, restoration becomes a significant challenge, with NED dropping sharply as fragmentation increases. Our findings highlight that current MLLMs lack the fine-grained cross-modal reasoning required to bridge visual discontinuities, identifying a critical gap in robust VRDU research.
LLM agents in interactive entertainment and social applications, including companions, role-play characters, and persona clones, need decisions grounded in personality and past experiences. We introduce HEART-Bench to evaluate memory-grounded persona-consistent response selection. The benchmark combines 11 Big Five-anchored synthetic characters, each with 1,000 autobiographical memories, and 64 DIAMONDS-guided decision scenarios into 673 multiple-choice questions. Domain experts independently compare candidate responses, with a senior expert resolving disagreements and revising responses when needed; other characters' adjudicated responses to the same scenario serve as distractors. In the main evaluation, models receive character IDs and occupations, scenarios, and options, with retrieved memories added in memory-augmented settings. Detailed personality-and-value profiles are withheld. We evaluate 16 LLMs under No-Retrieval, Naive-RAG, Mem0, and PersonaDB. Accuracy without memories stays near the 25% chance level, and the best memory-augmented configuration reaches 64.9%. On a matched 150-item subset, human annotators reach 80.8%, compared with 64.0% for the best model. Memory-content ablations show the importance of retrieving experiences that match both the target character and the decision context.
Model selection for imbalanced binary classification often uses the Matthews correlation coefficient (MCC), but its thresholding step makes validation rankings depend on the chosen cutoff. SoftMCC is a post-training MCC validation framework built on established probability-valued confusion counts, coupling an MCC-specific calibrated identity with a tie-aware, shared-pool selection protocol. Its core score is a covariance-normalized probability-label association, reduces exactly to MCC for hard predictions, and is Pearson-bounded. Under perfect population calibration it equals the Brier skill score with identical candidate ordering, which is the calibrated case of Murphy's partition of the Brier score. Outside that regime the gap does not identify calibration error. Across 18 settings with 12 duplicate-safe grouped repeats, SoftMCC attains the best stability mean rank (2.31) and highest mean tie-corrected Kendall's W (0.659), with a significant Friedman test (p=0.004). Nemenyi analysis separates it from area under the precision-recall curve (AUPRC) and MCC@0.5, and 14-source-family sensitivity retains both separations. For selected-model utility, none of six prespecified paired comparisons is significant before or after Holm correction (smallest raw p=0.093), and the dataset-level test is not significant (p=0.746). Label permutation lowers mean W to 0.092. Temperature scaling shifts SoftMCC rankings (mean Spearman 0.851) whereas rank-based and threshold-optimized metrics remain invariant. SoftMCC is a calibration-sensitive MCC-family selector with bounded stability evidence and inconclusive utility differences.
Large Reasoning Models (LRMs) generate long reasoning traces that expose intermediate inference behavior, offering opportunities to study and control reasoning in AI coding. However, the underlying patterns in these traces and their potential for analyzing and guiding coding models remain poorly understood, limiting systematic characterization and inference-time control of LRM reasoning. We conduct an empirical study of reasoning patterns across four representative coding tasks. We transform reasoning traces into structured action sequences and mine recurring fragments using within-task support, cross-task contrasts, and valid--invalid support contrasts. We characterize four general reasoning patterns by coding agents: design-to-implementation for code generation, state tracing for execution reasoning, evidence--fix--recheck for debugging, and semantic mapping for translation. We further identify positive and negative patterns associated with valid and invalid reasoning traces, respectively. Positive patterns reflect verification-driven convergence, whereas negative patterns reflect repeated design revision, reflection, and knowledge recall without closure. To demonstrate their practical utility, we evaluate three pattern-based applications: task identification, pattern-guided prompting, and pattern-aware early stopping. Features capturing general reasoning patterns enable accurate task identification from reasoning traces without accessing task descriptions or final answers. Prompting guided by positive and negative patterns improves the overall accuracy of downstream tasks, while negative pattern-aware early stopping substantially reduces API-reported token usage with only a modest decrease in pass@1. All three applications also generalize to previously unseen models, showing that reasoning patterns can support both the analysis and control of LRM behavior in AI coding.
Aijia Cheng, Kailong Wang Huazhong University of Science and Technology, Ling Shi, Yongxin Zhao East China Normal University
Function calling empowers large language models (LLMs) to interface with external tools, yet existing RL-based approaches suffer from misalignment between reasoning processes and tool-call decisions. We propose R2IF, a reasoning-aware RL framework for interpretable function calling, adopting a composite reward integrating format/correctness constraints, Chain-of-Thought Effectiveness Reward (CER), and Specification-Modification-Value (SMV) reward, optimized via GRPO. Experiments on BFCL/ACEBench show R2IF outperforms baselines by up to 34.62% (Llama3.2-3B on BFCL) with positive Average CoT Effectiveness (0.05 for Llama3.2-3B), enhancing both function-calling accuracy and interpretability for reliable tool-augmented LLM deployment.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
Large language models (LLMs) have achieved remarkable success across a wide range of natural language processing tasks, yet their performance remains heavily biased toward high-resource languages. Tibetan, despite its cultural significance and large speaker population, is still substantially underrepresented. In this work, we present a comprehensive pipeline for advancing Tibetan language modeling through large-scale data curation and continual pre-training. We construct a 72 GB high-quality Tibetan corpus, the largest to date, and adapt Qwen2.5-7B through balanced multilingual continual pre-training with Tibetan, Chinese, and English, followed by multilingual instruction tuning. To further scale capacity efficiently, we extend the dense model to a 50B-A10B Mixture-of-Experts architecture. Due to the absence of standardized Tibetan benchmarks, we build multiple evaluation datasets via high-quality translation and human verification. Experimental results show that both dense and MoE models consistently outperform existing open-source and Tibetan-focused models of similar scale across diverse tasks. Our work advances Tibetan-centric LLM research and provides transferable insights for extending LLMs to other low-resource languages. We will release the model weights, evaluation benchmarks, and detailed data processing documentation in the follow-up.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表
While mechanistic interpretability tools like Sparse Autoencoders (SAEs) can uncover meaningful features within Large Language Models (LLMs), a critical gap remains in transforming these insights into practical actions for model optimization. We bridge this gap with the hypothesis that data selection guided by a model’s internal task features is a effective training strategy. Inspired by this, we propose Interpretability-Guided Data Selection (IGDS), a framework that first identifies these causal task features through frequency recall and interventional filtering, then selects “Feature-Resonant Data” that maximally activates task features for fine-tuning. We validate IGDS on mathematical reasoning, summarization, and translation tasks within Gemma-2, LLaMA-3.1, and Qwen3 models. Our experiments demonstrate exceptional data efficiency: on the Math task, IGDS surpasses full-dataset fine-tuning by a remarkable 17.4% on Gemma-2-2B while using only 50% of the data, and outperforms established baselines focused on data quality and diversity. Analysis confirms a strong positive correlation between feature amplification and task performance improvement. IGDS thus provides a direct and effective framework to enhance LLMs by leveraging their internal mechanisms, validating our core hypothesis.
尚无 AI 解读，展示原文摘要
cs.CL
Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表