cs.CL · 2026-10-15 · No. 144

Computation and Language, 2026-10-15.

26 new papers in cs.CL. Titles, authors, abstracts. Links to arXiv. Want this in your inbox every morning? Subscribe →

01 — The papers

26 entries
  1. 01

    Latent Compaction with Test-Time Training

    Yingfa Chen, Xingyu Shen, Chaojun Xiao, Xu Han

    cs.CL · cs.LG

    Long contexts enlarge key--value (KV) caches, increasing memory use and attention costs during generation. Compressing these caches must preserve information for unseen questions. We introduce **Latent Compaction**, which learns a compact document-specific cache through test-time training. We initialize cache entries using reconstruction importance, teacher attention, and key-space diversity. Training first fits attention outputs and...

    arxiv.org/abs/2610.20582 · PDF

  2. 02

    PerSo: Investigating Facets of LLMs' Social Intelligence in Real-World Persuasion Scenarios

    Ali Zamani, Soroush Esfahanian, Erfan Shahabi, Mohammad Taher Pilehvar, Azadeh Shakery, Yadollah Yaghoobzadeh

    cs.CL · cs.AI · cs.SI

    Understanding the social intelligence of large language models (LLMs) is increasingly important for applications involving human-AI interaction and multi-agent environments. Social intelligence comprises two aspects: cognitive intelligence, the ability to infer others' mental states, and behavioral intelligence, the ability to select appropriate social actions based on these inferences to achieve social goals. Persuasion provides a natural...

    arxiv.org/abs/2610.20548 · PDF

  3. 03

    PrismaCoder: Pushing the Boundaries of Open Multimodal Code Intelligence through Decomposed Rubric Synthesis

    Qiushi Sun, Jingyang Gong, Yang Liu, Qiaosheng Chen, Zhangyue Yin, Zixian Huang, Hanxu Hu, Jing Xiong, Fangzhi Xu,...

    cs.CL · cs.AI · cs.CV · cs.LG · cs.SE

    Multimodal code intelligence connects the symbolic world of programs with the visual world their execution creates. Interest in this capability is burgeoning, as models are increasingly asked to produce charts, webpages, animations, and semi-structured documents. Recently, proprietary models have pushed far ahead, producing interactive artifacts even from a text description or a screenshot alone. Yet their data and training remain closed, and...

    arxiv.org/abs/2610.20538 · PDF

  4. 04

    VASPO: Self-Distillation with Privileged Value Estimation

    Fei Wu, Zhenrong Zhang, Jianshu Zhang, Jun Du

    cs.CL · cs.AI · cs.LG

    On-policy self-distillation provides dense supervision for reasoning models via a privileged self-teacher. However, this supervision reflects local token preferences rather than the expected return of an unguided student continuation. Recent methods combine this signal with GRPO. They use GRPO for update directions and the teacher for signal scaling. However, we demonstrate a more effective integration within PPO. Specifically, we identify...

    arxiv.org/abs/2610.20516 · PDF

  5. 05

    Lost in Translation: Why Lower Bias Scores Don't Mean Fairer Models

    Harsh Kumar

    cs.CL · cs.LG

    Bias audits of multilingual language models usually compare a model's average stereotype score in English with its average in another language and read the difference as a difference in fairness. We ask how much of that difference is noise. Auditing ten open models in eight languages, we measure how far scores move when English test items are merely reworded, test every language shift item by item, and check that the test catches shifts we...

    arxiv.org/abs/2610.20479 · PDF

  6. 06

    Equal CER, Unequal Damage: Severity-Weighted Character Error Rate for OCR

    Nguyen Nguyen, Yongyi Zang

    cs.CL · cs.AI · cs.CV

    Character Error Rate (CER), the standard metric for OCR and vision-language models (VLMs), charges a typo as much as a wrong name. We introduce semCER, which keeps CER's alignment and denominator and charges each grapheme edit a severity from an LLM judge, with no word segmentation. Normalized CER keeps the score of a whole transcription, where it tracks readers as well as semCER or better, and semCER breaks that score down into named edits....

    arxiv.org/abs/2610.20461 · PDF

  7. 07

    The COUNT: Counting N-grams Unlocks Scaling on Host Memory and Training-Free Adaptation

    Ian Magnusson, Noah A. Smith, Pang Wei Koh

    cs.CL · cs.LG

    The scaling relationship of compute and performance for language models (LMs) has been extensively studied; more recently, models are being designed to scale memory use, as well. In particular, adding embeddings for n-grams scales memory separately from compute use, because each pass only processes individual embedding rows, which are much smaller than the storage used for the whole embedding table. However, existing methods train embeddings...

    arxiv.org/abs/2610.20456 · PDF

  8. 08

    Analyzing performance of Retrieval Augmented Generation models for Question Answering Tasks

    Fazle Mohammed Tasfiq, Sabab Attin, Sheikh Md. Tahmid Hossain, Raiat Reza, Tanzim Ahmmed, Nafis Chowdhury, Anika...

    cs.CL · cs.AI

    Retrieval-Augmented Generation (RAG) models have become almost a cornerstone of open-domain question answering tasks; in which the quality of answers generated in the process, greatly relies on the effectiveness of retrieval mechanisms. In the realm of LLM, the last two years have seen tremendous progress in RAG architectures, with many models reporting to have improved performance through various design innovations. The cultural question and...

    arxiv.org/abs/2610.20419 · PDF

  9. 09

    From Unreliable to Usable: Decomposing Failure Modes and Confidence Signals in LLM Entity Extraction - Evidence from Cyber Threat Intelligence

    Jiahong Qian

    cs.CL · cs.AI

    Audits of large language model (LLM) extraction from full-length documents report severe unreliability, leaving two questions open: what the failures actually consist of, and whether confidence signals can turn unreliable output into usable output. We run a preregistered, fully frozen evaluation of two new-generation low-cost LLMs (DeepSeek-V4-Flash, GLM-5.3-Flash) with a flagship-tier control (GPT-6.1-Sol) on a public corpus of 354 incident...

    arxiv.org/abs/2610.20417 · PDF

  10. 10

    Cortex: Consolidating Context into Parametric Memory

    Jintian Zhang, Feng He, Xinnian Liang, Ziwen Xu, Haoming Xu, Haojing Huang, Shuangzhi Wu, Ningyu Zhang

    cs.CL · cs.AI · cs.CV · cs.LG · cs.MA

    Context can improve large language model performance, but these gains may be lost after context removal. Inspired by biological systems consolidation, we formulate Memory Consolidation (MC) as internalizing these gains while preserving non-target capabilities, connecting context-specified editing, skill learning, and behavioral steering under a common objective. We introduce a five-task-family testbed measuring consolidation and retention...

    arxiv.org/abs/2610.20399 · PDF

  11. 11

    FOLD: Position-Free Far Memory with Local Order for Long-Context Speculative Decoding

    Shuxuan Li, Jialiang Kang, Yanghao Zhou, Ziye Zhu, Xiao Yang

    cs.CL · cs.LG

    Speculative decoding accelerates LLM generation through lightweight drafting and parallel verification. It is particularly promising for long inputs, where KV-cache access costs can be amortized over multiple accepted tokens. Yet most drafters are trained on short sequences and degrade on long inputs, while long-context training is costly. Existing training-free approaches restrict the context window, rescale positions, or augment recent...

    arxiv.org/abs/2610.20392 · PDF

  12. 12

    MMLA-RSI: Recursive Self-Improvement through RTT and RTMA

    Junyi Zou, Avrova Donz

    cs.CL · cs.LG

    An agent can improve its behavior within an unresolved problem by changing how it acts and which experience it retains. We propose MMLA-RSI, a recursive self-improvement scheme with two implementations in Memory-Mediated Learning Architecture: Reasoning-Time Training (RTT) changes the policy state $Φ$, while Reasoning-Time Memory Adaptation (RTMA) changes authoritative memory $M$. Each route also makes its adaptation method an object of...

    arxiv.org/abs/2610.20294 · PDF

  13. 13

    CSWBench: Evaluating Context-Sensitive Code-Switching in Creative Writing

    Zewen Jiang, Yahan Li, Chenlu Guo, Diana Inkpen, Yuan Wu, Yi Chang

    cs.CL · cs.AI

    Large language models (LLMs) have demonstrated multilingual capabilities across diverse language generation and understanding tasks. Yet effective code-switching requires more than multilingual proficiency: models must decide when, where, and why to switch languages in ways that fit the surrounding context. This context-sensitive ability remains insufficiently evaluated. We introduce CSWBench, a benchmark for evaluating context-sensitive...

    arxiv.org/abs/2610.20272 · PDF

  14. 14

    Scaling Multilingual Mixture-of-Experts: How Language Coverage Changes Capacity and Compute Allocation

    Joel Dag, Nikit Srivastava, Hamada M. Zahera, Axel-Cyrille Ngonga Ngomo

    cs.CL · cs.LG

    Multilingual models share capacity across languages, so adding languages can hurt performance on existing ones. Mixture-of-experts models expand this capacity with little additional per-token computation. Given this extra capacity, does adding experts reduce what each additional language costs? To answer this question, we propose a repetition-aware joint scaling law relating number of active parameters, training tokens per language, number of...

    arxiv.org/abs/2610.20247 · PDF

  15. 15

    Scaling Multilingual Safety and Reasoning Evaluation through Reasoning-Guided Translation

    Zahraa Al Sahili, Ioannis Patras, Matthew Purver

    cs.CL · cs.AI

    Large language models (LLMs) are increasingly deployed across languages, yet their safety and reasoning capabilities remain predominantly evaluated in English. Translating existing benchmarks offers a scalable approach to multilingual evaluation, but preserving task-specific meaning across languages presents significant challenges, particularly for harmful requests, mathematical reasoning, and conversational interactions. We introduce a...

    arxiv.org/abs/2610.20221 · PDF

  16. 16

    What Makes a Social Agent's Interactions Effective? Learning from Verifiable Outcomes in Paired Worlds

    Caixin Kang, Ke Wang, Yuchuan Wu, Xiaoqiang Wang, Ling Li, Bingxue Qiu, Yongbin Li, Yoichi Sato

    cs.CL · cs.AI · cs.SI

    Effective social interaction requires more than fluent responses: agents must reach plans that fit the participants' needs and circumstances. Existing evaluations of social agents mainly rely on LLM judges that rate the conversation, without checking the resulting plan. We introduce HICO-Bench, a benchmark of 180 everyday coordination scenarios, such as resource allocation, scheduling, and authorization, in which two agents with private...

    arxiv.org/abs/2610.20195 · PDF

  17. 17

    Cross-Model In-Context Neurofeedback Reveals (Limited) Activation Control in LLMs

    Davide Baldelli, Robyn Kozierok, Mirko Bronzi, Damiano Fornasiere

    cs.CL · cs.AI · cs.LG

    Probes on activation vectors are widely used to detect misbehavior in language models, but their success relies on the assumption that models cannot control what their activations encode. Humans can learn, through neurofeedback, to control the firing of even single neurons, and recent studies investigated whether this ability extends to language models, reaching conflicting conclusions. No experimental design so far disentangles competence at...

    arxiv.org/abs/2610.20179 · PDF

  18. 18

    Gap Detection: Mitigating Homogenization through Demand and Defensibility

    Xucheng Yu, Haibo Jin, Xiaoqi Wang, Runze Kong, Haohan Wang

    cs.CL · cs.LG

    Language models often produce semantically similar answers to open-ended questions, even when many plausible alternatives exist. Generating useful alternatives requires identifying both what the query calls for and what existing responses already cover. We introduce Gap Detection, a framework that searches for relevant but unoccupied directions in a shared text embedding space. Its objective, GapScore, combines query relevance (Demand) with...

    arxiv.org/abs/2610.20162 · PDF

  19. 19

    AFCS: Atomic-Fact-Conditioned Synthesis for Factual Knowledge Pretraining

    Yunhao Wang, Feng Zhang, Tinghao Yu, Xing Wu, Jiang Zhou

    cs.CL · cs.AI

    Synthetic data is increasingly used in language-model pretraining. Existing knowledge-corpus synthesis pipelines, including WRAP and Nemotron-CC, usually take a source passage or document segment as input and generate rewritten or QA-style samples from that text. This can improve surface form and prompt diversity, but it does not explicitly require fine-grained source facts -- especially secondary entities, list/table entries, and precise...

    arxiv.org/abs/2610.20086 · PDF

  20. 20

    Which Knowledge Should LLMs Trust? Uncovering a Generalizable Steering Direction Across Knowledge Conflicts

    Yibin Lou, Yi Hu, Jianhui Chen, Chenpeng Wang, Muhan Zhang, Liangming Pan

    cs.CL · cs.AI

    Large Language Models (LLMs) frequently encounter conflicts between internal parametric knowledge and external contextual knowledge, yet the mechanisms governing this selection remain unclear. We show that the choice between context and parametric knowledge is linearly represented and can be controlled along a single direction in the model's residual stream. We first demonstrate that this direction generalizes across factual-conflict...

    arxiv.org/abs/2610.20067 · PDF

  21. 21

    A Strong Harness Narrows the Model Gap: A Comparative Evaluation for Legislative Document-to-XML Conversion

    Palvika Bansal, Rittika Jindal, Ruichong Wang

    cs.CL · cs.AI

    Legal-content reproduction requires matching the source exactly in section hierarchy, verbatim text, citations, and amendment spans; a dropped citation or mislabeled amendment carries into the products lawyers rely on. We report an industrial pipeline converting U.S. state legislation into strict, schema-constrained XML, and hypothesize that its accuracy comes from harness engineering, the deterministic tooling built around a model, more than...

    arxiv.org/abs/2610.19999 · PDF

  22. 22

    Choosing the Number of Attention Heads from Intrinsic Dimension: A Small-Scale Study

    Arjun Boya, Charan Sai Bandaru, Masthanaiah Pillimetla, Kavana Chidanandappa

    cs.CL · cs.LG

    The number of attention heads in a transformer is usually set by convention or grid search. We test a simple heuristic for setting it in advance: estimate the intrinsic dimension of token representations without labels, and use roughly that many heads. We estimate intrinsic dimension with a nested-dropout VAE trained on DistilBERT embeddings of WikiText-2, choosing the KL weight at the knee of the VAE's rate-distortion curve rather than by...

    arxiv.org/abs/2610.19966 · PDF

  23. 23

    AtomSearch: Evidence-Grounded Tree Search for Agent Skill Optimization

    Jia Tong, Senyou Deng, Min Peng, Jianfeng Pan

    cs.CL · cs.AI · cs.LG

    Agent skills adapt frozen models through editable external instructions. Revising several rules at once makes changes difficult to inspect, while judging each local edit only by its immediate score can discard useful intermediate states. We introduce AtomSearch, which keeps revision actions local while using subsequent skill states to guide search. Contrastive atomic editing links each primitive rule edit to failed and successful executions,...

    arxiv.org/abs/2610.19951 · PDF

  24. 24

    Does Machine Translation Retain Media Bias?

    Silvia V. Gargova, Tim Menzner, Jochen L. Leidner

    cs.CL · cs.AI

    Within the most recent two decades, machine translation (MT) has been advanced from a research area to a mature, deployed technology for prose text and most language pairs. However, language has different dimensions of meaning, and the alluded progress has primarily been demonstrated for propositional meaning. In this small exploratory study, we investigate whether one type of affective meaning, namely different types of linguistically...

    arxiv.org/abs/2610.19945 · PDF

  25. 25

    Frustratingly Easy Majority Voting Using Length Conditioning

    Fiorenzo Parascandolo, Enver Sangineto, Xiaoyi Yu, Wenhui Tan, Jipei He, Ruihua Song, Rita Cucchiara

    cs.CL · cs.AI

    Majority Voting (MV) is a simple and widely used strategy which improves the reasoning accuracy of Large Language Models (LLMs) by allocating additional inference-time computation. Specifically, in MV multiple Chain-of-Thoughts (CoTs) are independently sampled and the most frequent final answer is selected. In this work, we propose Length-Conditioned Majority Voting (LCMV), a Test-Time Training approach which improves MV leveraging an...

    arxiv.org/abs/2610.19890 · PDF

  26. 26

    Rationale Distribution Distillation via Flow Matching for Distributional Continuous Latent Reasoning

    Mingyeong Kim, Gyusang Cho, Juho Lee, Giung Nam

    cs.CL · cs.AI

    Sampling multiple chain-of-thought rationales from a language model can uncover correct solutions missed by a single attempt. However, these rationales are typically lengthy, making repeated autoregressive generation costly. Continuous latent reasoning offers a compact alternative, but point-valued methods condition repeated answer sampling on a single rationale, which can limit exploration of diverse paths and answer coverage. We thus...

    arxiv.org/abs/2610.19871 · PDF

This edition is part of The Daily Abstract — cs.CL archive. Subscribe to receive these in your inbox each morning, automatically translated to Spanish, with reply-to-PDF: arxivdaily.ignorelist.com.

Colophon Set in Georgia, with system sans for interface chrome and a monospaced stack for code and paper identifiers. Sole accent: amber #D99C5E. Built and served on an always-free VM. The masthead is set 14% letterspaced because newspapers do that and it works.