Peer Revenue Publish or profit

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Computer Science Computation and Language

cs.CL Watch Computation and Language

Saturday 8 August

Computation and Language · Cross-listed from cs.LG

Spectral Outliers Reveal Dominant Learned Structure in Transformer Attention

Kasun Dewage, Marianna Pensky, Suranadi De Silva et al. (+1 more)

We apply Marchenko-Pastur (MP) random matrix theory to pre-trained attention weights in order to separate each projection matrix into a random-like bulk and a set of spectral outliers. We validate this decomposition causally: zeroing the MP-identified outliers (signal) in Mistral-7B drives HellaSwag, MMLU, and PIQA close to random-chance performance, whereas zeroing a count-matched subset of bulk singular values causes smaller but non-negligible degradation. Across 11 pre-trained transformers we identify five recurring patterns: spectral outliers encode a dominant component of the learned structure; Q projections carry the most outliers; V projections under grouped-query attention lack a clean signal/noise separation; entry-level outliers form structured row-bands in Q and column-bands in O; and specific residual-stream dimensions persist as band outliers across layers in K and O. We close by outlining how these observations could inform parameter-efficient fine-tuning and structured pruning.
Read the full abstract

All authors Kasun Dewage, Marianna Pensky, Suranadi De Silva, T. H. Bandara

Computation and Language

Detection of Self-Introductions in Legislative Testimony

Sofija Dimitrijevic, Pallavi Das, Kasey Liu et al. (+1 more)

Self-introductions are common in legislative committee testimonies. Successfully detecting them and extracting the speaker's name is enormously helpful in the task of speaker identification in the context of government meetings. In this paper, we present a pipeline for detection of self-introductions in legislative committee testimony using machine learning. We construct a training dataset from 1.54 million utterances spanning five state legislative sessions, apply a name-matching heuristic to generate automatic labels, and train three classifiers: a decision tree, random forest, and XGBoost to find self-introductions and extract the speaker's name. We construct a feature set combining bag-of-words, positional context, structural signals, introductory phrase indicators, and discourse context features. Among the three classifiers, XGBoost achieves the best performance with an F1 score of 0.9747 and the fewest total errors; adding fine-tuned BERT probability features improves this further. As an extension, we score the full candidate dataset with a fine-tuned BERT classifier and add BERT probability outputs as features. This BERT-augmented XGBoost model improves F1 from 0.9747 to 0.9782 and reduces total test errors from 241 to 207. The primary gain over the decision tree baseline (F1 0.9323) is driven by discourse context features and the boosting ensemble strategy; BERT provides a modest complementary signal. Analysis of false positives reveals that a minority are genuine self-introductions mislabeled due to name inconsistencies in the source data, indicating that measured metrics modestly understate true performance.
Read the full abstract

All authors Sofija Dimitrijevic, Pallavi Das, Kasey Liu, Foaad Khosmood

Computation and Language · Cross-listed from cs.CV

Vision-Language Grounding as Bidirectional Concept Correspondence

Jieyu Zhang, Ziqi Gao, Luke Zettlemoyer et al. (+1 more)

Vision-language grounding connects language to visual content, yet most existing formulations reduce grounding to a unidirectional localization problem: given a prespecified text phrase or category name, identify the corresponding image region. This setup assumes that the relevant linguistic unit is already known, overlooking a more basic challenge in grounded communication: determining which parts of the text are visually referential and how they correspond to entities in the image. We formulate grounding as $\textit{bidirectional concept correspondence}$ over an image-text pair. Given an image and its paired text, the goal is to recover all correspondences between visually referential text spans and instance-level image segments, without assuming that the relevant text spans are provided. This formulation unifies common grounding tasks, including phrase grounding, referring expression grounding, and open-vocabulary detection, by treating text segmentation, image segmentation, and cross-modal alignment as a single correspondence prediction problem. To address this task, we introduce $\textbf{ConCor-1}$, a grounding model built on top of a pretrained vision-language model. It uses learnable $\textit{bridge tokens}$ to represent candidate image-text correspondences and predicts, for each token, a text mask, an image mask, and a correspondence presence score. To train and evaluate this task, we convert diverse grounding and segmentation datasets into a unified correspondence format. Experiments show that $\textbf{ConCor-1}$ consistently outperforms baselines, improving correspondence F1 by 48% on the long-caption dataset and by 29% on zero-shot LVIS, where the large category list serves as the text input.
Read the full abstract

All authors Jieyu Zhang, Ziqi Gao, Luke Zettlemoyer, Ranjay Krishna

Computation and Language · Cross-listed from cs.AI

GRACE: LLM-Grounded Semantic Metric Spaces for Scalable Mixed-Data Clustering

Zihua Yang, Zhencheng Xie, Junyang Chen et al. (+4 more)

Clustering mixed tabular data requires a unified metric space to bridge the inherent heterogeneity between continuous numerical measurements and discrete categorical symbols. Traditionally, algorithms rely entirely on dataset-internal statistics to estimate categorical relationships, which confines the learned metric to empirical co-occurrences and ignores conceptually obvious yet statistically unobserved affinities. Although LLMs offer external world knowledge, applying their text-centric reasoning to highly abstract tabular concepts presents significant challenges. Bridging this modality gap to construct a semantically complete metric typically requires embedding LLMs into iterative metric learning loops to dynamically optimize cross-modality representations. This incurs intractable computational overhead, forcing a compromise between semantic enrichment and scalability. Therefore, we propose GRACE, an LLM-grounded framework for scalable mixed-data clustering. GRACE shifts semantic acquisition to the attribute-value level via a multi-perspective LLM querying strategy, mapping heterogeneous values into knowledge-informed descriptions. Crucially, this one-shot grounding extracts general-purpose semantic representations that embed heterogeneous attributes into a unified space, decoupling expensive LLM invocation from iterative optimization. Furthermore, GRACE cross-validates these external semantics against dataset-internal statistical evidence to ensure alignment with the dataset-specific cluster structure. Ultimately, GRACE matches the scalability of conventional statistics-driven baselines while achieving superior clustering accuracy and conceptual interpretability over 11 competing methods. The source code is available at https://github.com/develop-yang/GRACE-GRACE-A
Read the full abstract

All authors Zihua Yang, Zhencheng Xie, Junyang Chen, Liang Xie, Yiqun Zhang, Mengke Li, Yang Lu

Computation and Language

SurakshaEval: An Indic Safety Benchmark for Multilingual LLMs

Debopriyo Banerjee, Kapil Rajesh Kavitha, Angana Borah et al. (+11 more)

Existing safety evaluation datasets for large language models (LLMs) predominantly focus on English and Western contexts, often overlooking the linguistic diversity and culturally grounded safety risks present in other languages. To address this gap, we introduce SurakshaEval, a novel safety benchmark composed of human-written prompts spanning real-world scenarios, explicitly designed for ten major Indian languages - Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Punjabi, Tamil, and Telugu, along with English. SurakshaEval includes both generic prompts common across India and region- and language-specific prompts that capture localized sociocultural sensitivities. We benchmark a broad range of state-of-the-art LLMs on SurakshaEval, establish baseline safety performance, and identify recurring failure modes, including over-refusal, missed detection of implicit bias, and insufficient contextual awareness in regionally sensitive settings. Our results show that even strong multilingual LLMs struggle to reliably meet nuanced safety requirements when operating in Indic languages, particularly in native scripts. These findings highlight the urgent need for safety evaluation frameworks that incorporate region-specific data and structured assessment protocols, enabling the development and deployment of AI systems that operate securely, ethically, and in alignment with diverse societal values. Our code and data are available at https://github.com/debobanerjee/SurakshaEval. Warning: This paper contains text that may be offensive or unsafe.
Read the full abstract

All authors Debopriyo Banerjee, Kapil Rajesh Kavitha, Angana Borah, Xudong Han, Yuxia Wang, Parameswari Krishnamurthy, Utkarsh Agarwal, Atharva Kulkarni, Swaran Lata, Ayush Munot, Dhruv Sahnan, Aaryamonvikram Singh, Preslav Nakov, Monojit Choudhury

Computation and Language

"Many Are My Names": The Anatomy of the Assistant and Its Personas via Sparse Autoencoders

Adelaide Danilov, Aria Nourbakhsh, Oleksandr Marchenko Breneur et al. (+1 more)

How a language model internally represents who is speaking, the Assistant, an assigned roleplay persona, or a narrated story character, remains underexplored. We study speaker representations using a dataset of user-expressed emotional text and corresponding model responses. We decompose three generation settings (Assistant, Roleplay, and Story) into sparse autoencoder features extracted at turn-boundary and pronoun-token positions and selected through a filtering pipeline for different depths. We characterize each surviving feature through its steering effects and activation distribution. Our main finding is that the Assistant and roleplay personas are not independent alternatives: personas retain the Assistant-associated feature core while progressively differentiating from it across layers, starting from operational machinery towards behavioral and stylistic features. Meanwhile, generated story characters lack the Assistant-associated core. Both Story and Roleplay can be distinguished from the Assistant with Immersive Simulation Mode. However, the Assistant can sometimes enter or slowly drift into it even in the default setting.
Read the full abstract

All authors Adelaide Danilov, Aria Nourbakhsh, Oleksandr Marchenko Breneur, Salima Lamsiyah

Computation and Language · Cross-listed from cs.LG

TEMPER: Tensorized Efficient Manifold-constrained Parameterization for Expressive Residual Routing

Yuxuan Gu, Wuyang Zhou, Huijun Xing et al. (+1 more)

Residual connections rely on a static residual pathway, and are essential for training deep neural networks. Hyper-connections (HC) increase the expressivity of residual routing by incorporating multiple residual streams and learning dynamic information flow, while manifold-constrained (mHC) variants stabilize training through doubly stochastic residual mixing. However, a generator-level bottleneck remains in existing methods: they use dense, unstructured generators for pre-branch aggregation, residual mixing, and post-branch redistribution, which results in parameter count growing rapidly with the number of streams. To address this issue, we propose \underline{T}ensorized \underline{E}fficient \underline{M}anifold-constrained \underline{P}arameterization for \underline{E}xpressive Residual \underline{R}outing (TEMPER), which represents these generators as multi-way tensors over the input-stream, feature, and output-stream modes, and parameterizes them using tensor networks. Such a structured low-rank formulation is shown to preserve token-dependent manifold-constrained routing interface while substantially reducing parameter growth. It also promotes interpretability and intuition, as: i) tensor ranks control the dimensionality of the learned routing subspace, with full ranks recovering dense routing; while ii) the generator approximation errors bound differences in routing logits and, consequently, in the routed-block outputs. Comprehensive experiments show that TEMPER matches or outperforms existing methods across language modeling and commonsense reasoning tasks, while requiring substantially fewer additional parameters. At eight residual streams, TEMPER achieves the best CORE score while using about $84\%$ fewer additional parameters than mHC, thus showing a stronger performance-parameter efficiency trade-off.
Read the full abstract

All authors Yuxuan Gu, Wuyang Zhou, Huijun Xing, Danilo Mandic

Computation and Language · Cross-listed from cs.LG

From token probabilities to calibrated confidence: An empirical study of mathematical question answering

Avery Ma, Lorne Schell, Vin Bhaskara et al. (+1 more)

Confidence estimation for large language models (LLMs) aims to estimate the probability that a generated answer is correct, while calibration aligns these estimates with empirical accuracy. Prior work has shown that token probabilities are often overconfident, we investigate whether these readily available signals can nevertheless provide well-calibrated confidence estimation for mathematical question answering. We compare single-pass estimators, which reuse token probabilities from the original generation, with multi-pass estimators, which obtain additional confidence signals through verification or stochastic forward passes. While individual token probabilities can be highly saturated, we find that aggregating token probabilities over the full sequence captures small but consistent differences between correct and incorrect generations, yielding more informative confidence estimates. Multi-pass methods can yield calibrated confidence estimates. We study two such approaches: self-verification through re-prompting, including a lower-cost in-situ variant, and Monte Carlo Dropout, which derives confidence from variation across stochastic forward passes. We further evaluate two post-hoc calibration methods, Platt scaling and isotonic regression, both of which substantially reduce in-domain calibration error. However, their data efficiency varies with dataset difficulty, and the calibration mappings often transfer asymmetrically across datasets and models.
Read the full abstract

All authors Avery Ma, Lorne Schell, Vin Bhaskara, Leila Pishdad

Friday 7 August

Computation and Language

Enhancing Audio Reasoning via Semantic Summary Prediction

Francesco Bonzi, Pooneh Mousavi, Cem Subakan et al. (+1 more)

Large Audio Language Models (LALMs) perform well on complex question answering but often show a reasoning gap, where explicit Chain-of-Thought (CoT) reduces accuracy compared to direct answers. We hypothesize that long reasoning sequences shift attention away from the audio input. To address this, we propose SPARE (Semantic Prediction for Audio REasoning), which introduces a register token aligned with the final conclusion using a cosine similarity loss with a Sentence-BERT embedding. This conditions the model's latent space with the target semantic goal before reasoning begins. Experiments on MMAU and MMAR with SALMONN show improved zero-shot reasoning and stronger early attention to audio without additional inference cost.
Read the full abstract

All authors Francesco Bonzi, Pooneh Mousavi, Cem Subakan, Mirco Ravanelli

Computation and Language

PAUSE: Editable Strategy Artifacts for Long-Form Cultural Story Adaptation

Taaha Kazi, Vasu Sharma, Mohammad Saifullah et al. (+1 more)

Generative AI systems increasingly mediate cultural adaptation, but their cultural decisions are often hidden inside prompts, transient model plans, or final prose. We study PAUSE (Pause-And-Update Strategy Editing), an intervention that exposes an editable adaptation strategy as a human control surface for cultural decisions in long-form story adaptation. The strategy is a structured artifact that can be inspected, edited, and then projected through downstream character, entity, and chapter-localization stages. In two Chinese-source serialized novels, we test whether human edits to this strategy propagate into chapter-level prose. Across 9 edited-vs-control chapter comparisons, judges select the edited-strategy output in all 9; a marker audit shows target markers in 8/9 edited outputs and 0/9 controls, with forbidden markers absent from edited outputs and present in all controls. We frame these results as a smoke-scale edit-adherence study, not a claim that the outputs are culturally authoritative or literary-quality improvements. PAUSE offers one practical way to make AI-mediated cultural adaptation more inspectable and contestable before decisions propagate through long-form generation.
Read the full abstract

All authors Taaha Kazi, Vasu Sharma, Mohammad Saifullah, Abdur Rahman

Computation and Language · Cross-listed from cs.CY

Generative Gap Filling

Yonathan A. Arbel and David A. Hoffman

Most contract litigation turns on contracts that imperfectly record parties' bargains. When the parties' dispute can't be solved by interpreting the text, courts fill the gap. Scholars have long assumed that the remaining text runs out quickly, and provides thin evidence of the actual deal on the disputed point. On that view, a judge who supplies the missing term must be drawing on something else, from commercial defaults to her own policy preferences. Despite generations of work, courts have no real alternative to such unruly methods. We tested that assumption. Taking real contracts, we masked a term the parties had negotiated and asked readers to predict what we removed. Lay respondents recovered the hidden term about half the time, twice what chance predicts. Law students and lawyers did marginally better. But large language models, given nothing but the rest of the contract, recovered it nearly nine times in ten. The deal, in short, testifies to far more of the agreement than the literature assumes, including terms the parties never wrote. A contract, we argue, is like a radio signal from far away. Even when incomplete, enough of the message is carried elsewhere that the missing part can be reconstructed with the right receiver. True gaps are rarer than supposed. Courts can weigh model predictions as ordinary, contestable evidence, and parties can discipline the practice with "Choice of Model" clauses.
Read the full abstract

Computation and Language · Cross-listed from cs.IR

Do LLM Recommenders Know When They're Hallucinating? Auditing Confidence Calibration in Catalog Faithfulness

Srijith Ravikumar

LLM recommenders for top-K item suggestion regularly emit titles outside the target catalog. Prior audits report a binary out-of-domain rate; none ask whether the model knew. We jointly audit hallucination rate (OOD@10) and verbalized-confidence calibration (ECE, Brier, reliability) for four zero-shot LLM recommenders from four independent vendors (Mistral Large, Llama-3.3-70B, GPT-OSS-120B, Claude Sonnet 4.6), not grounded or fine-tuned systems, across three catalogs (MovieLens-25M, Amazon Reviews 2023 Toys, Yelp Open Dataset), stratified by item popularity. Measuring catalog membership is itself the hard part: on identical outputs the reported rate moves by an order of magnitude with the string matcher used, and F1 cannot separate the candidates. We validate the instrument against 201 human judgments and select on net bias, where the adopted one is off by -0.040 against +0.144 for the common fuzzy rule. Hallucination is then strongly catalog-dependent (0.6-2.7% on MovieLens, 11.6-38.7% on Yelp, 49.3-61.0% on Amazon Toys). Each model holds a near-constant confidence level barely responsive to the catalog, while the catalog-hit rate swings 60 points, so the sign of the error is set by where a model's constant lands against a catalog's accuracy: 7 of the twelve cells are under-confident and 5 over-confident, all four under-confident on MovieLens, all four over-confident on Amazon Toys. We read this as an elicitation mismatch: "Just Ask" elicits a generic quality rating, not a catalog-membership probability. A conformal abstention threshold over verbalized confidence changes hallucination by at most 1.65 pp across four alpha levels, because the channel cannot separate correct items from hallucinations. We recommend that audits report calibration alongside OOD, validate the matcher producing the OOD number, and use catalog-anchored elicitation.
Read the full abstract

2608.10008 · 7 Aug · Accepted at CIKM 2026 (short paper). v2 corrects the catalog-membership matcher, which scored truncated catalog entries as perfect matches to longer hallucinated titles. One finding reverses: v1 reported all four models under-confident in all twelve cells; v2 finds the catalog sets the direction. Amazon Toys hallucination moves from 4.5-8.3% to 49.3-61.0%. Cite v2 or later; v1 is superseded

Abstract PDF $HALU ×4$AUDIT ×1$CELL ×1$DATA ×1

Computation and Language

On the use of foundation models in cognitive science

Raj Sanjay Shah, Alex Warstadt, Michael Frank et al. (+1 more)

A host of recent studies have evaluated the cognitive and developmental alignment of Foundation Models (FMs). These investigations include evaluations of their correspondence to adult performance across a range of cognitive domains, as well as whether aspects of model training track children's cognitive development. However, using FMs as candidate cognitive models poses significant methodological and conceptual challenges. A key question underlies this effort: under what conditions does behavioral alignment justify treating FMs as explanatory models of cognition? In this paper, we articulate a four-stage inferential framework for evaluating FMs as cognitive and developmental models: adapting human experimental tasks to model-compatible formats, specifying linking hypotheses that map model outputs to human measures, evaluating behavioral correspondence, and comparing across candidate models or manipulations. We clarify the role of linking hypotheses in mapping model outputs to human behavioral measures, identify challenges that constrain alignment claims, and propose principles for theory-driven and comparative evaluation. Throughout, we argue that behavioral fit alone is insufficient. Alignment becomes scientifically meaningful only when embedded within explicit theoretical commitments, theory-diagnostic tasks, and systematic contrastive evaluation across candidate models.
Read the full abstract

All authors Raj Sanjay Shah, Alex Warstadt, Michael Frank, Sashank Varma

Computation and Language

Jako Tako or Fluent? Presenting PoVisLE: A Polish Vision-Language Evaluation

Anna Kołos, Grzegorz Statkiewicz, Karolina Seweryn et al. (+3 more)

Vision-language models (VLMs) have achieved strong performance on tasks such as image captioning, visual question answering, and image-to-text generation. However, they are predominantly trained on English-centric data, which limits their ability to handle culturally grounded visual understanding and leads to failures in interpreting region-specific meanings, symbolic content, and context-dependent visual cues. Existing benchmarks for cultural competence are often template-driven and focused on surface-level recognition, making them insufficient for evaluating deeper linguistic and pragmatic understanding in culturally situated settings. We introduce PoVisLE, a monocultural vision-language benchmark for Polish designed to evaluate culturally grounded multimodal understanding under a grounded evaluation paradigm, where language is interpreted in interaction with visual context. The dataset contains 1,117 images and 2,366 manually annotated VQA pairs. Overall, our dataset provides a controlled and challenging resource for assessing culturally grounded vision-language understanding beyond surface-level recognition.
Read the full abstract

All authors Anna Kołos, Grzegorz Statkiewicz, Karolina Seweryn, Katarzyna Kowol, Karolina Piosek, Wojciech Kusa

Computation and Language

The No-Meaning Falsity: The Structural Impossibility of the Arbitrary Sign in Classical Arabic

Elnaserledinellah Mahmoud Abdelwahab

This paper investigates whether the postmodern claim of unrestricted semantic indeterminacy, and its foundational Saussurean axiom of the arbitrary sign, are compatible with the structural architecture of Classical Arabic. We develop a formal mathematical model of Arabic non concatenative morphology in which lexical meaning is determined by the interaction between an invariant root and a morphosyntactic pattern. Within this framework, we establish a Morphological Correspondence Theorem, demonstrating that every lexical item is uniquely generated by a root pattern pair, and a Semantic Localization Theorem, proving that lexical meaning is determined at the derivational level prior to surface realization. To address Saussurean weaker notion of relative arbitrariness, we formalize it via conditional Kolmogorov complexity, defining arbitrariness algorithmically as the no rule property. We prove that general relative arbitrariness is formally undecidable, while Arabic relative arbitrariness is decidable and provably less than 1 for its motivated signifiers (Levels W and M), establishing a strict system complexity asymmetry over Indo-European languages.
Read the full abstract

Computation and Language

Evaluating Dedicated Monolingual and Joint Multilingual Causal Models for Dravidian Languages

Venkata Naga Sai Vishnu Rohit Pulipaka

Dravidian languages, mainly Tamil, Telugu, Kannada, and Malayalam make up only a small part of the data used to train multilingual language models, so it's not clear how much per-language ability these models actually keep. I have trained five GPT-2 architecture models from scratch to compare four monolingual models (one each for Tamil, Telugu, Kannada, and Malayalam, each with its own 32K-vocabulary subword tokenizer) against one multilingual model sharing a 64K-vocabulary subword tokenizer across all four languages. All the 5 models are trained on cleaned CC-100, Wikipedia, and Samanantar data. I have tested the models on perplexity, bits-per-byte, tokenizer efficiency, and fine-tuning results which are compared against mGPT. The monolingual models outperform mGPT on sentiment classification and named entity recognition, and their tokenizers proved more efficient than the shared multilingual model across all the languages tested.
Read the full abstract

Computation and Language · Cross-listed from cs.CV

CosmosAlign: Adapting a World Foundation Model for Generative Traffic Video Forecasting

Quang Minh Dinh and Tuan Kiet Doan

Generative traffic video forecasting aims to synthesize long-horizon, temporally coherent future videos of traffic scenes from a short observation history and textual descriptions. In this paper, we present CosmosAlign, a generative traffic video forecasting framework built upon the pretrained Cosmos3-Nano world foundation model. Our approach is motivated by the observation that successfully adapting large pretrained world models to downstream forecasting tasks depends primarily on distribution alignment rather than increased model capacity. To this end, we propose a two-stage LoRA adaptation strategy that first aligns the conditioning-mode distribution with the target forecasting task, and then aligns the training captions with the model's native structured prompting interface through an LLM-based re-captioning pipeline. During inference, we further improve prediction quality using a fully training-free procedure consisting of consensus-based medoid sample selection and motion-adaptive blending of static scene regions. CosmosAlign achieves a final score of 76.49 on the AI City Challenge 2026 Track 5 benchmark, ranking first on the final leaderboard. Our code is publicly available at https://quangminhdinh.github.io/CosmosAlign/.
Read the full abstract

Computation and Language · Cross-listed from cs.AI

IntelliAudit: Using Large Language Models to Evaluate Audit Controls

Allison Wilson, Sina Moradi Sabet, Diar Shakimov et al. (+4 more)

IT audits require auditors to judge whether heterogeneous organizational evidence satisfies semantic security and compliance controls. This judgment is difficult to automate because relevant evidence is distributed across policies, records, spreadsheets, and operational artifacts, and because audit conclusions depend on evidentiary sufficiency rather than keyword matching. We present IntelliAudit, a retrieval-grounded multi-agent system for IT audit evidence evaluation. Given a control and an evidence corpus, IntelliAudit retrieves relevant artifacts, generates an evidence-grounded assessment, challenges adverse findings, adjudicates disagreements, and produces an auditor-facing recommendation with cited evidence, rationale, missing-evidence analysis, and remediation guidance. We instantiate IntelliAudit on ISO/IEC 27001 and evaluate it across multiple simulated organizations using expert auditor review and audit-readiness user feedback. The evaluation shows that IntelliAudit can support control interpretation, evidence-grounded reasoning, and audit-preparation workflows, while also revealing the importance of human oversight for calibrating sufficiency judgments and correcting overly permissive recommendations. These results suggest that retrieval-grounded multi-agent systems can assist audit evidence review, but should remain decision-support tools rather than autonomous certification systems.
Read the full abstract

All authors Allison Wilson, Sina Moradi Sabet, Diar Shakimov, Panteha Shahrivar, Mohammad Reza Bagheri, Dean Konenkamp, Mohammad A. Tayebi

Computation and Language · Cross-listed from cs.CV

Keep It Simple: Multi-Key Episodic Memory Retrieval for Ultra-Long Video Understanding

Yeeun Choi, Youngbeom Yoo, Joon-Young Lee et al. (+2 more)

When videos extend from hours to days, directly processing them end-to-end becomes impractical for current Multi-modal Large Language Models (MLLMs). This ultra-long setting necessitates a two-stage paradigm: query-agnostic memory construction followed by retrieval-based inference. Prior work invests in complex memory construction to pre-model high-level relations in videos, despite not knowing the downstream query at build time. We instead prioritize high-recall retrievability during memory building, and defer query-specific, high-level relation composition to inference time. To this end, we propose MERIT(Multi-key Episodic Retrieval with Inference-time Temporal expansion), a simple yet effective agentic framework for ultra-long video understanding. First, we formulate an episodic multi-key representation that enables precise retrieval of fine-grained memories through a simple key-matching mechanism. Second, we introduce a neighbor filtering mechanism to capture broader semantic context without the massive computational overhead of global memory construction. This is achieved by expanding the temporal scope exclusively around the retrieved segments at inference time. By leveraging simple key-matching with this on-demand temporal expansion, MERIT achieves state-of-the-art performance across three long-video benchmarks: EgoLifeQA, LVBench, and Video-MME (Long).
Read the full abstract

All authors Yeeun Choi, Youngbeom Yoo, Joon-Young Lee, Hyolim Kang, Seon Joo Kim

Computation and Language

SurveyReview: A Reviewer-Aligned Benchmark for Survey Evaluators

Yuheng Zhang, Yuanchun Wang, Fanjin Zhang et al. (+4 more)

The rapid advancement of large language models has transformed survey writing from a months-long manual effort into an automated process. As generation scales, reliable evaluation becomes the bottleneck, and LLMs are increasingly used as survey evaluators. However, existing approaches largely rely on off-the-shelf LLM-as-a-judge methods without systematic alignment to human reviewers, and there remains a lack of systematic frameworks for quantifying alignment with human reviewers. To address this gap, we propose SurveyReview, a reviewer-aligned, multi-dimensional benchmark and dataset for survey evaluation. We collect and annotate 675 survey papers with 1,630 review reports. We structure authentic peer-review reports by converting free-form comments into four-dimensional scores (Readability, Criticalness, Comprehensiveness, Structure) paired with supporting rationales. We further release standardized train/test splits and an evaluation protocol to measure alignment between automatic evaluators and human reviewers. To validate the benchmark, we develop SurveyAlign, a strong baseline evaluator by fine-tuning Qwen3-32B with LoRA on our annotated data, augmented with external knowledge for knowledge-intensive dimensions. On the test set, SurveyAlign substantially improves reviewer alignment over prompt-based judging with GPT-5.2, reducing average MSE from 2.28 to 1.38 and MAE from 1.15 to 0.69 across all four dimensions. Our contributions are twofold: (1) we establish the first multi-dimensional, reviewer-aligned dataset with a reproducible evaluation framework for survey reviewing; (2) we develop a strong baseline evaluator that substantially improves alignment with human reviewers, providing a competitive reference for future research. Our code and data are available at https://surveyreview.github.io
Read the full abstract

All authors Yuheng Zhang, Yuanchun Wang, Fanjin Zhang, Ruyu Zhao, Juanzi Li, Jie Tang, Jing Zhang

Computation and Language

Embedding Initialization for Unseen Low-resource Languages in Multilingual NMT: A Case Study on Limbum-English Translation

Samiratu Ntohsi, Neza David Tuyishimire, Anesu Kafesu et al. (+3 more)

Multilingual neural machine translation models such as NLLB-200 cover 200 languages but leave thousands unsupported, including most Grassfields Bantu languages of Cameroon. When fine-tuning these models for an unseen language, practitioners must choose a proxy language token, yet no principled method exists for this selection. We implemented an embedding initialization strategy where a language token is the average of embeddings from multiple typologically related languages already in the mod el. We evaluate this approach on Limbum-to-English translation using a parallel corpus of 8,837 sentence pairs from New Testament text and a bilingual dictionary. We compare models: NLLB-200 zero-shot (chrF2++ = 12.5), a Transformer trained from scratch (chrF2++ = 14.5), NLLB-200 fine-tuned with a Swahili proxy token (chrF2++ = 47.3), and NLLB-200 with our averaged embedding initialization (chrF2++ = 46.7). We find that the multi-language initialization achieves performance comparable to the best single-language proxy. Both NLLB-200 variants improve over the from-scratch baseline by over 32 chrF2++ points. These results show that multilingual transfer is the dominant factor in extremely low-resource Bantu translation while eliminating the need for heuristic proxy selection. However, all systems fail to preserve tonal diacritics, highlighting an open challenge. We make our dataset and code available to support further research.
Read the full abstract

All authors Samiratu Ntohsi, Neza David Tuyishimire, Anesu Kafesu, Marvin Ogore, Samuel Oluwajunwonlo Babalola, Oche Ankeli

Computation and Language · Cross-listed from cs.AI

From Single Chatbots to Governed Agent Ecosystems: An Agentic AI Pattern Catalogue and Orchestration Framework for Mission-Critical Hospital Information Management Systems

Manideep Dhar, Ritwik Singh and Sharat Chandra Kumar Manikonda

Hospitals are racing to embed AI, while coping with the surge in adaptation of the technology in other industries, into the triage management, documentation, scheduling, and revenue-cycle workflows, yet most deployments remain as fragmented pilots that stall at the edge of production, exposing patients and institutions to operational fragility, ungoverned risk, and mounting technical debt. At the same time, the global AI-in-healthcare market is projected to exceed nearly USD 1 trillion by 2034, according to the report of Fortune Business Insights, amplifying the financial consequences of architectural missteps and failed scaling strategies. This research proposes a compliance-first Agentic AI pattern catalogue and orchestration framework, purposely built for HIMS, moving beyond the single LLM chatbots and towards a governed ecosystem of autonomous and semi-autonomous agents. The framework extends by adding (i) a taxonomy of Agentic roles, (ii) a formal risk-stratification model that maps each pattern to risk tiers, human-in-the-loop checkpoints, and governance hooks, and (iii) a unified orchestration runtime capable of coordinating multi-agent workflows across EHR/HIMS landscapes such as Epic, Cerner, and MEDITECH. Technically the framework combines vLLM-based inference, optimized paging memory, confidential computing, and MCP based on-premise deployment, enforcing end-to-end encryption and policy-as-code controls aligned with HIPAA, GDPR, the EU AI Act, India's DPDP and DISHA Acts, ISO 27001, ISO 27002, ISO 14971 and IEC 62304. We exhibit how the proposed architecture is capable and efficient to reduce the documentation time, integration effort, and AI pilot attrition while constricting the governance and auditability, offering hospital leaders and governing authorities an urgently needed blueprint to convert AI investment into sustainable clinical, operational, and financial ROI
Read the full abstract

Computation and Language · Cross-listed from cs.SE

DevIntent: How Much Does LLM-Generated Code Violate Developer Intent?

Susana Haing, Natan Vidra and Spurthi Setty

Code generated by LLMs can violate a developer's implicit intentions when given an ambiguous prompt, yet standard benchmarks measure only whether code passes its stated test. We introduce the Intent Violation Rate (IVR) and a 49-problem pilot benchmark derived from HumanEval+. Each problem strips implicit constraints from a clarified prompt and encodes them as hidden constraint tests. IVR measures the fraction of LLM-generated solutions that pass the stated (visible) tests yet fail hidden constraint tests that capture unstated intent. Evaluating Claude Sonnet 4.6 and OpenAI GPT 4.1, we find both pass over 92% of stated tests yet violate intent in over half of problems (54.5% and 63.5%), following a systematic, bimodal pattern consistent across both models. Out findings indicate that pass rates overstate how well generated code reflects developer intent.
Read the full abstract

Computation and Language

CreativeInstruct: Scalably Teaching LLMs to Balance Quality, Creativity, and Diversity

Ananya Sahu, Mohit Bansal and Elias Stengel-Eskin

While post-training improves the capabilities of large language models (LLMs), it generally lowers their output diversity and creativity, negatively impacting tasks that explicitly require creativity (e.g., story generation) as well as those that require it implicitly, e.g., reinforcement learning (RL). We instead propose CreativeInstruct, a scalable instruction-tuning method that teaches LLMs to balance creative, base-model-like generations with the quality of post-trained models, by learning to inject special [StartCreativity] spans that bias generation toward creativity. Furthermore, we introduce a structural diversity metric based on graph edit distance, which captures narrative level variation missed by purely lexical and semantic metrics. On narrative generation, CreativeInstruct matches or exceeds the diversity of both multi-model baselines and distilled variants of their outputs, without sacrificing quality or requiring multiple models at inference time. These results are mirrored in our human evaluation, where we find that annotators rate CreativeInstruct generations as more creative than the post-trained LLMs' generations in 70.3% of cases. We also show the benefits of creative models as a substrate for RL: GRPO applied to a CreativeInstruct checkpoint improves by ~4% on AMC and ~5% points on MATH over the same training applied to the post-trained checkpoint.
Read the full abstract

Computation and Language

CoinRAG: Contextualized Information Nugget KV Cache Reuse for Long-Context RAG

Gyuwan Kim, Cheoneum Park and Tao Yang

Recent optimization studies on Retrieval-Augmented Generation (RAG) have exploited chunk-level KV cache reuse to avoid processing long retrieved contexts for higher efficiency, while significant information redundancy and noise still remain in the coarse-grained chunks. This paper optimizes the Pareto frontier under low prefill latency constraints while maximizing accuracy by proposing CoinRAG (Contextualized Information Nugget KV Cache Reuse for Long-Context RAG). The name metaphorically reflects our core mechanism: much like assembling small tokens (or "coins") to accumulate a larger value, CoinRAG compositionally reuses offline-computed, fine-grained nugget caches to form a learned contextual representation efficiently in a more semantically relevant but compact manner. Specifically, instead of full-chunk encoding, CoinRAG identifies query-relevant semantic units within retrieved chunks through two-stage retrieval and seamlessly assembles their sliced KV representations with a chunk-level context. Extensive evaluations on LongBench multi-hop question answering tasks demonstrate that CoinRAG significantly reduces operational costs and outperforms the other baselines with a new Pareto frontier and an average 5.3% relative improvement in answer quality (F1) under a standard fast prefill latency budget.
Read the full abstract

Computation and Language · Cross-listed from cs.AI

SkillProx: Self-Evolving Agent Skills via Proximal Textual Gradient Descent

Mingxuan Zheng, Yujin Zhou, Chuxue Cao et al. (+6 more)

LLM agents increasingly adapt to recurring tasks by accumulating procedural knowledge in skills. These skills are lightweight, reusable textual artifacts that are loaded into the agent's context without weight updates. Recent methods refine skills through iterative task execution, failure diagnosis, and trajectory-guided text-space updates. However, existing frameworks lack explicit diagnosis–outcome feedback and treat deletion as a generic edit operation rather than a dedicated mechanism for consolidating accumulated knowledge. We introduce SkillProx, a proximal-gradient-inspired forward–backward framework that couples closed-loop diagnostic evolution with utility-aware proximal refinement. Motivated by a composite objective balancing task loss and skill complexity, the forward stage re-executes diagnosis-driven edits on the same task batch, rolls back regressions, and feeds measured outcomes into subsequent diagnoses. The backward stage decomposes the resulting skill into auditable knowledge units, estimates their contributions using a frozen leave-one-out utility audit, and applies validation-gated consolidation, demotion, or removal. Experiments on in-distribution and out-of-distribution benchmarks across multiple backbone LLMs show that SkillProx improves average accuracy by 3.0 percentage points over the strongest gradient-based baseline. Component ablations demonstrate the complementary effects of closed-loop diagnosis and proximal refinement.
Read the full abstract

All authors Mingxuan Zheng, Yujin Zhou, Chuxue Cao, Boqin Yin, Yuyao Zhang, Jiapeng Sun, Shuaishuai Gong, Sirui Han, Yike Guo

Computation and Language

An Exploratory Evaluation of LLM-Assisted Rewriting of Moderate-Complexity Financial Sentences for DisCoCat-Based Sentiment Analysis

Brian Llinas and Nikos Chrisochoides

Quantum natural language processing (QNLP) provides a grammar-aware framework for text modeling, and Distributional Compositional Categorical (DisCoCat) is one of its theoretically grounded formulations. Prior work on financial sentiment analysis has identified practical limitations of DisCoCat, including parser sensitivity, high simulation cost, and difficulty handling longer sentences. We study an LLM-assisted preprocessing workflow that uses controlled rewriting to compress, simplify, or decompose moderate-complexity financial sentiment sentences into parser-compatible, circuit-efficient variants while preserving sentiment-bearing meaning. We compare prompting strategies, language models, and filtering configurations with the low-complexity-only DisCoCat baseline of Stein et al. At the circuit level, the strongest compression variants reduce average qubit and gate counts by more than 70 percent relative to the raw moderate-complexity subset. Across repeated training runs, GPT-4.1-mini with Prompt B achieves the highest observed mean accuracy, $0.550 \pm 0.035$, compared with $0.521 \pm 0.050$ for the baseline. Larger training splits do not necessarily improve downstream performance; across evaluated configurations, training-split size has a moderately negative association with accuracy (Pearson $r=-0.446$). These results provide exploratory evidence that LLM-assisted rewriting can make some moderate-complexity inputs usable within the evaluated DisCoCat configuration, while highlighting prompt design, filtering, and circuit-aware preprocessing as considerations for more scalable QNLP-based financial sentiment analysis.
Read the full abstract

Computation and Language · Cross-listed from cs.AI

PsychoAgent: An Affect-Sensitive Cognitive Architecture for Conflict-Aware Memory in LLM Agents

Mohammad Amanlou, Parham Abed Azad, Farbod Davoodi et al. (+3 more)

Human-like cognition does not select past experience by topical similarity alone: affective significance and unresolved conflict also shape what becomes accessible. We present PsychoAgent, a cognitive architecture for LLM agents that separates factual and affective memory and integrates both through a conflict-aware executive controller. Affective memories are first filtered by semantic relevance and then re-ranked by salience, preserving topical fit while allowing emotionally important traces to enter the prompt. Across three controlled conflict scenarios, the full architecture retrieved more conflict-critical memories than semantic-affective and single-memory RAG baselines (0.933 vs. 0.500 and 0.667), with a small semantic-similarity cost. Five blinded raters evaluated 27 outputs. After within-rater standardization, the full architecture had the highest overall mean (+0.22 SD), but corrected pairwise differences were not significant. A three-day illustrative trace further shows persistent affect, offline memory recombination, and selective memory reweighting. The findings support affect-sensitive retrieval as an inspectable mechanism for modeling human-like conflict effects in LLM agents.
Read the full abstract

All authors Mohammad Amanlou, Parham Abed Azad, Farbod Davoodi, Mostafa Masumi, Behnam Bahrak, Abdol-Hossein Vahabie

Computation and Language · Cross-listed from cs.CV

SABRE: Scalable and Automated Benchmarking of VLMs under Stress

Zixuan Lan, Luzhe Sun, Matthew R. Walter et al. (+1 more)

Vision-language models (VLMs) are improving rapidly, but benchmark development lags behind, making weaknesses hard to identify. Building stress tests is costly: samples must satisfy controlled conditions, remain answerable, and challenge current models. We present SABRE, a scalable, automated pipeline that converts a Test Primer (a Markdown Task Design with Data Schema) into structured specifications, generated or edited images, and question-answer pairs. Automated filtering removes candidates solved by a Filtering VLM, while human review verifies candidate validity and supports annotation correction and localized image repair. We instantiate SABRE-Prior to test whether VLMs follow visual evidence instead of relying on world priors – learned expectations about familiar objects and scenes. Its 600 images and 1,000 questions span Context (unexpected entities in familiar scenes), Texture (counterfactual materials), Attribute (noncanonical component counts), and Language Elicitation (answers suggested by language but unsupported by the image). Across six VLMs, macro-average accuracy ranges from 17.8% to 31.3% (22.6% mean). A real-image Attribute control is comparably difficult for the Filtering VLM. SABRE-Counting and SABRE-Spatial pilots show that the workflow supports other stress-test settings. These results establish SABRE as a reusable framework for constructing and refreshing VLM stress tests rather than a single fixed benchmark.
Read the full abstract

All authors Zixuan Lan, Luzhe Sun, Matthew R. Walter, Jiawei Zhou

Computation and Language · Cross-listed from cs.AI

ResidencyRL: Reinforcement Learning in Simulated Clinical Environments

Valentin Liévin, Samuel Schmidgall, Tim Strother et al. (+32 more)

In medical education, physicians convert academic knowledge into clinical expertise through residency: years of training across thousands of encounters, with diverse sources of feedback and progressively greater autonomy. Much of clinical reasoning relies on the patient encounter, a dialogue in which a clinician elicits history, refines diagnostic hypotheses, and decides management under uncertainty. While large language models (LLMs) excel on static medical benchmarks, methods to optimize the full sequence of clinical decisions remain underdeveloped. We present ResidencyRL, a reinforcement learning (RL) method for training clinical artificial intelligence (AI) agents through simulated multi-turn clinical encounters (up to 60 dialogue turns and 8 tool calls per trajectory). ResidencyRL pairs the policy agent with LLM simulators capable of complex, adversarial behaviors, training against a structured reward aligned to diagnostic accuracy, management quality, communication, documentation, and safety. On held-out evaluations, the ResidencyRL agent improves diagnostic accuracy by 7.0% under adversarial conditions (88.0% vs. 81.0%) and reduces missed red flag rates by 31%, demonstrating rigorous mitigation of premature closure. Blinded expert clinicians validated these gains, preferring the trained agent in 87.6% of side-by-side comparisons. The procedural competencies transfer to unseen benchmarks: the agent outperforms the base model across all six clinical axes of the AMIE multi-visit benchmark, and shows consistent directional improvements on AgentClinic and CRAFT-MD. Our findings demonstrate that sequential clinical decision-making can be effectively learned through multi-turn RL in simulation, yielding robust, generalizable capabilities, paving the way towards clinical mastery. Prospective validation with real-world workflows remains necessary to establish clinical utility.
Read the full abstract

All authors Valentin Liévin, Samuel Schmidgall, Tim Strother, Alex Bijamov, Akshay Goel, Anil Palepu, Chunjong Park, Vahid Balazadeh, Min Woo Sun, Marius Guerard, Justin Chen, Dave Steiner, Vikram Dhillon, Ibrahim Azar, Akhil Mehta, Nicholas Spetsieris, Shilpan Shah, Maen Abdelrahim, Amit Dahiya, Yun Liu, Katherine Chou, Yossi Matias, Avinatan Hassidim, Dale R. Webster, Quoc V. Le, Raia Hadsell, Joelle Barral, Carey Radebaugh, Aleksandra Faust, Shekoofeh Azizi, Mike Schaekermann, Po-Hsuan Cameron Chen, Tao Tu, David Racz, Lin Yang