cs.CL · 2026-10-07 · No. 136

Computation and Language, 2026-10-07.

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

01 — The papers

18 entries
  1. 01

    IdeaAnchor: Teaching LLMs to Turn Literature into Research Ideas

    Ziyu Chen, Yilun Zhao, Jiashuo Sun, Yiling Ma, Manasi Patwardhan, Arman Cohan

    cs.CL · cs.AI

    Scientific research often begins by synthesizing ideas from a set of related papers to identify gaps and formulate new directions. However, training language models to perform this form of literature-grounded ideation remains challenging, as existing approaches based on prompting or feedback lack structured supervision for how papers should be synthesized. We introduce IdeaAnchor, a paradigm for training LLMs to perform research ideation...

    arxiv.org/abs/2610.08781 · PDF

  2. 02

    AdvSim2Real : Training Web Agents Against Adaptive Prompt Injection in a Web World Model

    Sarim Hashmi, Mukul Ranjan, Kshitij Mishra, Mikhail Kuznetsov, Praneeth Vepakomma, Nils Lukas

    cs.CL · cs.AI · cs.LG

    Web agents complete user requests by reading and acting on pages that third parties write, so an instruction planted on a page can redirect the agent away from the user's goal. The agent cannot simply ignore the page, because the page also holds the values and controls the task requires. Current defenses fine-tune the agent on injections fixed before training, and attackers that adapt to the trained model bypass them. Adversarial training...

    arxiv.org/abs/2610.08773 · PDF

  3. 03

    Denoising Hierarchical Representations: Joint Continuous Diffusion for Language Modeling

    Mathias Ollu, Nikos Komodakis

    cs.CL · cs.LG

    Diffusion Language Models (DLMs) hold the promise of order-agnostic, parallel text generation. Recently, continuous diffusion and flow matching models have seen substantial gains, driven by carefully crafted token representations and diffusion/flow spaces. In this work, we introduce Hierarchical Continuous Diffusion Language Models (H-CDLMs), a simple framework that further improves continuous DLMs with minimal compute and parameter overhead....

    arxiv.org/abs/2610.08738 · PDF

  4. 04

    When Forgetting is not Catastrophic: On the Mechanics of Spurious Forgetting

    Vedant Palit, Florent Draye, Nicolas Zucchet, Zhijing Jin, Bernhard Schölkopf

    cs.CL · cs.LG

    Knowledge that a language model appears to forget during finetuning often remains stored and can be recovered, a phenomenon called spurious forgetting. Finetuning on new facts can even produce forgetting that undoes itself: recall of the old facts collapses, recovers as training continues on new facts alone, and only then erodes for good. We seek to understand when such forgetting is not catastrophic. A minimal associative memory reproduces...

    arxiv.org/abs/2610.08718 · PDF

  5. 05

    Latent space bias directions in LLMs capture confidence, not fairness

    Stephanie Buttigieg, Maeve Madigan, Parameswaran Kamalaruban, Stuart Burrell

    cs.CL · cs.AI · cs.LG

    Activation steering has gained popularity as a lightweight inference-time debiasing technique for large language models. However, prior work reports that steering vectors generalise poorly, with unintended effects on model performance and limited transfer to new datasets. Our work analyses what the debiasing direction used for activation steering actually encodes, in order to shed light on its inconsistent performance. We study the linear...

    arxiv.org/abs/2610.08559 · PDF

  6. 06

    How High Is 0.6? Floors, Ceilings, and Headroom in Interpretability Probing

    Pranjal Garg

    cs.CL · cs.AI

    Probes are the workhorse of interpretability. If a model's hidden states predict a variable, the model is said to represent it. But a probe score has no fixed meaning. An $R^2$ of 0.6 may only reflect what the input already gives away, and the same score can mean different things on different data. We propose reading every probe score against two reference points: a floor, what a declared set of simple inputs already predicts, and a ceiling,...

    arxiv.org/abs/2610.08544 · PDF

  7. 07

    Wiki-Talkie: Multilingual Benchmarking of Persona-Based Agents on Real-World Discussions

    Dennis Fucci, Andrea Bacciu, Dong Liu, Weronika Łajewska, Saab Mansour

    cs.CL · cs.AI

    LLMs are increasingly deployed as autonomous agents in social environments, making it critical to study their ability to faithfully simulate human interactions. Central to this is grounding agents in realistic user personas, yet existing datasets rely on fictional personas and are limited to a handful of languages, lacking the empirical grounding necessary to evaluate behavioral fidelity across diverse populations. We introduce Wiki-Talkie, a...

    arxiv.org/abs/2610.08513 · PDF

  8. 08

    Language-model ratings of depression reflect the rater more than the patient

    Baihan Lin

    cs.CL · cs.AI · q-bio.NC

    Depression has no diagnostic blood test. Language models promise tireless, consistent assessment, but can accurate raters disagree about individuals? We pre-registered 880 language-model raters, crossing 11 open models with prompting and scoring choices, and applied them to 189 interviews against the eight-item Patient Health Questionnaire. Model choice explained 30.0% of summed-symptom score variance, stable participant differences 10.5%....

    arxiv.org/abs/2610.08501 · PDF

  9. 09

    UNREAL: Unifying Retrieval and Long-Context with a Single Model

    Edan Kinderman, Elad Hoffer, Yochai Blau, Brian Chmiel, Ron Banner, Daniel Soudry, Boris Ginsburg

    cs.CL · cs.IR · cs.LG

    Long-context inference and Retrieval-Augmented Generation (RAG) handle evidence selection at vastly different scales, from a single long prompt to an entire corpus. We ask whether a single model-internal mechanism can select evidence across this range. We introduce UNifying REtrieval And Long-Context with a Single Model (UNREAL), a model-native evidence selection framework to span corpus retrieval and long-context inference. UNREAL encodes...

    arxiv.org/abs/2610.08463 · PDF

  10. 10

    Agentic AutoRAG: RAG Pipeline Optimization through Reasoning-Driven Agents

    Lasse B. Strand, Robert Jakob, Kevin O'Sullivan, Markus Kreft

    cs.CL · cs.IR · cs.LG

    Retrieval-augmented generation (RAG) is a widely used approach for grounding large language models (LLMs) in external knowledge. However, configuring a pipeline is an expensive hyperparameter optimization problem over many interacting choices, from chunking and embedding model to reranking and generation. Existing optimizers, from greedy search to Bayesian optimization, reduce each trial to an aggregate score and search without modeling why a...

    arxiv.org/abs/2610.08452 · PDF

  11. 11

    Rethinking Cross-Tokenizer On-Policy Distillation: From Alignment Coverage to Supervision Reliability

    Bingxi Hou, Guochao Jiang, Guofeng Quan, Weiqing Li, Wenfeng Feng, Guohua Liu, Yuewei Zhang

    cs.CL · cs.AI

    On-Policy Distillation (OPD) trains a student on its own generations using teacher feedback. With different tokenizers, comparing teacher and student predictions requires alignment at both sequence and vocabulary levels. In this paper, we examine whether expanding this alignment coverage improves learning. Across three heterogeneous teacher--student pairs on mathematical reasoning and code generation, strict 1:1 groups already cover most...

    arxiv.org/abs/2610.08448 · PDF

  12. 12

    Knowing When Not to Answer: Cross-Domain and Multi-Turn Generalization of Latent Underspecification Signals

    Jerzy Kamiński, Ilya Galyukshev, Artem Kuznetsov, Danil Fedorov, Kirill Redko, Sergey Chuprin, Aidar Shumbalov,...

    cs.CL · cs.AI · cs.LG

    Large language models routinely answer questions that cannot be answered from the information given, and in dialogue they answer before enough has been said. Unanswerability is linearly decodable from hidden states, but it is unclear which of its forms share a representation and whether the signal is useful in dialogue. We contribute a turn-labeled multi-turn benchmark (423 conversations, 1,661 labeled turn-states) and an evaluation harness...

    arxiv.org/abs/2610.08413 · PDF

  13. 13

    Foresight-over-Graph: Reasoning Beyond Local Horizons for Knowledge Base Question Answering

    Yang Hong, Yajun Yang, Xin Wang, Liping Jing, Qinghua Hu

    cs.CL · cs.AI

    Large language models (LLMs) have demonstrated strong capabilities in question answering, yet they still frequently suffer from hallucinations on knowledge-intensive tasks. Knowledge graphs (KGs) provide LLMs with structured, interpretable, and updatable factual grounding, making them a promising external knowledge source for reliable reasoning. However, existing LLM-guided graph reasoning methods typically rely on hop-wise greedy or...

    arxiv.org/abs/2610.08388 · PDF

  14. 14

    STRUCTURALCOST: A controlled reading time dataset for modeling human sentence processing difficulty

    Nina Nusbaumer, Iria de-Dios-Flores, Corentin Bel, Christophe Pallier, Guillaume Wisniewski, Benoît Crabbé

    cs.CL · cs.AI

    We introduce STRUCTURALCOST, a self-paced reading dataset of 475 participants and 40,800 observations isolating the processing cost of long-distance subject-verb dependency resolution. We replicate a low-powered psycholinguistic finding at NLP scale, namely that human reading times at the main verb increase with dependency length, driven by syntactic embedding beyond linear distance. Different language models -- spanning n-gram models, SSMs,...

    arxiv.org/abs/2610.08208 · PDF

  15. 15

    Making COMET Comparable Across Scripts: Diagnosis and Correction of Tokeniser-Induced Script Bias in Indic MT Evaluation

    G. L. John Salvin, Swapnil Hingmire

    cs.CL · cs.LG

    COMET reports translation quality as a single number, and that number is routinely compared across target languages written in different scripts. Such a comparison assumes Script Invariance: the score should not depend on the writing system that carries the target. We test it on IndicMT Eval by re-encoding the target into Latin script, which changes orthographic form while holding content and human ratings fixed. Script identity then accounts...

    arxiv.org/abs/2610.08159 · PDF

  16. 16

    Penalty-Framed No-Valid-Option MCQA: Analyzing LLM Abstention under Invalid Choices

    Jinhyeok Kim, Hye-Young Jung

    cs.CL · cs.AI

    Multiple-choice question answering (MCQA) is commonly used to evaluate large language models under the assumption that one of the provided options is correct, typically using answer-selection accuracy. However, in real deployments, users or retrieval systems may provide invalid option sets in which none of the listed choices is correct, and selecting one of them may incur downstream cost. We study this setting as penalty-framed...

    arxiv.org/abs/2610.08153 · PDF

  17. 17

    SAGE: Semantic Anchor-Guided Evolution for Grounded Medical QA Data Synthesis

    Chuan Li, Chengyu Wang, Cen Chen, Ye Lyu, Mingyuan Fan, Ming Gao

    cs.CL · cs.AI

    Developing reliable models for clinical tasks, such as Medical Question Answering (QA), is severely constrained by the limited availability of high-quality, expert-annotated training data. This challenge is exacerbated by stringent privacy requirements and the impracticality of utilizing large open-source corpora or proprietary cloud APIs within resource-limited clinical settings. To address these obstacles, we introduce SAGE...

    arxiv.org/abs/2610.08093 · PDF

  18. 18

    Language Carries the Expert's Impression: Instrument-Anchored LLM Judges Transfer Counseling-Quality Assessment and Beat In-Domain Training

    Tobias Hallmen, Elisabeth André

    cs.CL · cs.LG

    Automatic assessment of communication quality in dyadic counseling conversations is bottlenecked by data: expert-rated corpora are small and expensive to grow. We study cross-domain transfer of expert overall-impression prediction across three German corpora of simulated counseling (two general-practice medical, one school-related parent-teacher; $n=195$ expert-rated sessions, one corpus after scale equating). Training on the other domains...

    arxiv.org/abs/2610.08055 · 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.