Peer Revenue Past citations are no guarantee of future returns

Papers

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

Computer Science Computation and Language

cs.CL Watch Computation and Language

Tuesday 17 February

Computation and Language

Measuring Social Integration Through Participation: Categorizing Organizations and Leisure Activities in the Displaced Karelians Interview Archive using LLMs

Joonatan Laato, Veera Schroderus, Jenna Kanerva et al. (+3 more)

Digitized historical archives make it possible to study everyday social life on a large scale, but the information extracted directly from text often does not directly allow one to answer the research questions posed by historians or sociologists in a quantitative manner. We address this problem in a large collection of Finnish World War II Karelian evacuee family interviews. Prior work extracted more than 350K mentions of leisure time activities and organizational memberships from these interviews, yielding 71K unique activity and organization names – far too many to analyze directly. We develop a categorization framework that captures key aspects of participation (the kind of activity/organization, how social it typically is, how regularly it happens, and how physically demanding it is). We annotate a gold-standard set to allow for a reliable evaluation, and then test whether large language models can apply the same schema at scale. Using a simple voting approach across multiple model runs, we find that an open-weight LLM can closely match expert judgments. Finally, we apply the method to label the 350K entities, producing a structured resource for downstream studies of social integration and related outcomes.
Read the full abstract

All authors Joonatan Laato, Veera Schroderus, Jenna Kanerva, Jenni Kauppi, Virpi Lummaa, Filip Ginter

Computation and Language · Cross-listed from cs.AI

World-Model-Augmented Web Agents with Action Correction

Zhouzhou Shen, Xueyu Hu, Xiyun Li et al. (+3 more)

Web agents based on large language models have demonstrated promising capability in automating web tasks. However, current web agents struggle to reason out sensible actions due to the limitations of predicting environment changes, and might not possess comprehensive awareness of execution risks, prematurely performing risky actions that cause losses and lead to task failure. To address these challenges, we propose WAC, a web agent that integrates model collaboration, consequence simulation, and feedback-driven action refinement. To overcome the cognitive isolation of individual models, we introduce a multi-agent collaboration process that enables an action model to consult a world model as a web-environment expert for strategic guidance; the action model then grounds these suggestions into executable actions, leveraging prior knowledge of environmental state transition dynamics to enhance candidate action proposal. To achieve risk-aware resilient task execution, we introduce a two-stage deduction chain. A world model, specialized in environmental state transitions, simulates action outcomes, which a judge model then scrutinizes to trigger action corrective feedback when necessary. Experiments show that WAC achieves absolute gains of 1.8% on VisualWebArena and 1.3% on Online-Mind2Web.
Read the full abstract

All authors Zhouzhou Shen, Xueyu Hu, Xiyun Li, Tianqing Fang, Juncheng Li, Shengyu Zhang

Computation and Language

Vision Wormhole: Latent-Space Communication in Heterogeneous Multi-Agent Systems

Xiaoze Liu, Ruowang Zhang, Weichen Yu et al. (+7 more)

Heterogeneous multi-agent systems combine models with different capabilities through a common communication interface. Exchanging internal states directly requires translating between model-specific representations and controlling intermediate computation. We introduce the Vision Wormhole, which repurposes the visual input interface of Vision-Language Models (VLMs) for continuous communication between frozen heterogeneous agents. A Universal Visual Codec encodes each sender's latent rollout into a fixed-size message, maps it through a shared reference space, and decodes received messages into the receiver's image-token span. Per-model codecs and affine reference maps form a hub-and-spoke architecture with $O(N)$ components for $N$ models. Each model learns its codec independently through self-distillation on anchor texts, and shared-anchor alignment enables reuse across communication partners. Across four VLM families, six team configurations, and nine reasoning benchmarks, Vision Wormhole improves accuracy by 6.0 percentage points on average over text-mediated MAS and achieves a 1.69$\times$ geometric-mean speedup in batch-normalized end-to-end runtime.
Read the full abstract

All authors Xiaoze Liu, Ruowang Zhang, Weichen Yu, Siheng Xiong, Liu He, Feijie Wu, Hoin Jung, Matt Fredrikson, Xiaoqian Wang, Jing Gao

Computation and Language

Making Large Language Models Speak Tulu: Structured Prompting for an Extremely Low-Resource Language

Prathamesh Devadiga and Paras Chopra

Can large language models converse in languages virtually absent from their training data? We investigate this question through a case study on Tulu, a Dravidian language with over 2 million speakers but minimal digital presence. Rather than fine-tuning an LLM, we examine whether structured prompts alone can elicit basic conversational ability under controlled prompting. We systematically tackle various challenges posed by absence of training data for Tulu by combining explicit grammar documentation, negative constraints to suppress high-probability tokens from related languages, romanization standardization, and quality-controlled synthetic data generation via self-play. Evaluated on a manually curated held-out set across three LLMs (Gemini 2.0 Flash, GPT-4o, Llama 3.1 70B) and validated by native speakers, our approach reduces vocabulary contamination from 80% to 5% while achieving 85% grammatical accuracy. Cross-model analysis reveals that negative constraints provide consistent improvements (12–18 percentage points), while grammar documentation effects vary by model architecture (8–22 points).
Read the full abstract

Computation and Language

Orchestration-Free Customer Service Automation: A Privacy-Preserving and Flowchart-Guided Framework

Mengze Hong, Chen Jason Zhang, Zichang Guo et al. (+3 more)

Customer service automation has seen growing demand within digital transformation. Existing approaches either rely on modular system designs with extensive agent orchestration or employ over-simplified instruction schemas, providing limited guidance and poor generalizability. This paper introduces an orchestration-free framework using Task-Oriented Flowcharts (TOFs) to enable end-to-end automation without manual intervention. We first define the components and evaluation metrics for TOFs, then formalize a cost-efficient flowchart construction algorithm to abstract procedural knowledge from service dialogues. We emphasize local deployment of small language models and propose decentralized distillation with flowcharts to mitigate data scarcity and privacy issues in model training. Extensive experiments validate the effectiveness in various service tasks, with superior quantitative and application performance compared to strong baselines and market products. By releasing a web-based system demonstration with case studies, we aim to promote streamlined creation of future service automation.
Read the full abstract

All authors Mengze Hong, Chen Jason Zhang, Zichang Guo, Hanlin Gu, Di Jiang, Li Qing

Computation and Language

Far Out: Evaluating Language Models on Slang in Australian and Indian English

Deniz Kaya Dilsiz, Dipankar Srirag and Aditya Joshi

Language models exhibit systematic performance gaps when processing text in non-standard language varieties, yet their ability to comprehend variety-specific slang remains underexplored for several languages. We present a comprehensive evaluation of slang awareness in Indian English (en-IN) and Australian English (en-AU) across seven state-of-the-art language models. We construct two complementary datasets: WEB, containing 377 web-sourced usage examples from Urban Dictionary, and GEN, featuring 1,492 synthetically generated usages of these slang terms, across diverse scenarios. We assess language models on three tasks: target word prediction (TWP), guided target word prediction (TWP$^*$) and target word selection (TWS). Our results reveal four key findings: (1) Higher average model performance TWS versus TWP and TWP$^*$, with average accuracy score increasing from 0.03 to 0.49 respectively (2) Stronger average model performance on WEB versus GEN datasets, with average similarity score increasing by 0.03 and 0.05 across TWP and TWP$^*$ tasks respectively (3) en-IN tasks outperform en-AU when averaged across all models and datasets, with TWS demonstrating the largest disparity, increasing average accuracy from 0.44 to 0.54. These findings underscore fundamental asymmetries between generative and discriminative competencies for variety-specific language, particularly in the context of slang expressions despite being in a technologically rich language such as English.
Read the full abstract

Computation and Language

NeuroSymActive: Differentiable Neural-Symbolic Reasoning with Active Exploration for Knowledge Graph Question Answering

Rong Fu, Yang Li, Zeyu Zhang et al. (+7 more)

Large pretrained language models and neural reasoning systems have advanced many natural language tasks, yet they remain challenged by knowledge-intensive queries that require precise, structured multi-hop inference. Knowledge graphs provide a compact symbolic substrate for factual grounding, but integrating graph structure with neural models is nontrivial: naively embedding graph facts into prompts leads to inefficiency and fragility, while purely symbolic or search-heavy approaches can be costly in retrievals and lack gradient-based refinement. We introduce NeuroSymActive, a modular framework that combines a differentiable neural-symbolic reasoning layer with an active, value-guided exploration controller for Knowledge Graph Question Answering. The method couples soft-unification style symbolic modules with a neural path evaluator and a Monte-Carlo style exploration policy that prioritizes high-value path expansions. Empirical results on standard KGQA benchmarks show that NeuroSymActive attains strong answer accuracy while reducing the number of expensive graph lookups and model calls compared to common retrieval-augmented baselines.
Read the full abstract

All authors Rong Fu, Yang Li, Zeyu Zhang, Jiekai Wu, Yaohua Liu, Shuaishuai Cao, Yangchen Zeng, Yuhang Zhang, Xiaojing Du, Simon Fong

Computation and Language · Cross-listed from cs.LG

Discovering Implicit Large Language Model Alignment Objectives

Edward Chen, Sanmi Koyejo and Carlos Guestrin

Large language model (LLM) alignment relies on complex reward signals that often obscure the specific behaviors being incentivized, creating critical risks of misalignment and reward hacking. Existing interpretation methods typically rely on pre-defined rubrics, risking the omission of "unknown unknowns", or fail to identify objectives that comprehensively cover and are causal to the model behavior. To address these limitations, we introduce Obj-Disco, a framework that automatically decomposes an alignment reward signal into a sparse, weighted combination of human-interpretable natural language objectives. Our approach utilizes an iterative greedy algorithm to analyze behavioral changes across training checkpoints, identifying and validating candidate objectives that best explain the residual reward signal. Extensive evaluations across diverse tasks, model sizes, and alignment algorithms demonstrate the framework's robustness. Experiments with popular open-source reward models show that the framework consistently captures > 90% of reward behavior, a finding further corroborated by human evaluation. Additionally, a case study on alignment with an open-source reward model reveals that Obj-Disco can successfully identify latent misaligned incentives that emerge alongside intended behaviors. Our work provides a crucial tool for uncovering the implicit objectives in LLM alignment, paving the way for more transparent and safer AI development.
Read the full abstract

Computation and Language · Cross-listed from cs.LG

Prescriptive Scaling Reveals the Evolution of Language Model Capabilities

Hanlin Zhang, Jikai Jin, Vasilis Syrgkanis et al. (+1 more)

Machine learning model performance improvements tend to arise from competition and application. For deployment, we consider prescriptive scaling laws: given a pre-training compute budget, what downstream accuracy is attainable with contemporary post-training practice, and how stable is that mapping as the field evolves? Using large-scale observational evaluations with 5k existing and 2k newly evaluated model checkpoints spanning 2022-2026 across six benchmarks, we estimate capability boundaries, high conditional quantiles of benchmark scores as a function of log pre-training FLOPs, via smoothed quantile regression with a monotone, saturating sigmoid parameterization. We validate temporal reliability by fitting on earlier model generations and evaluating on later releases: across four of six tasks, the out-of-distribution coverage error remains below 2%, while math reasoning exhibits a consistently advancing boundary over time. For instance, at a budget of 10^24 FLOPs, the estimated attainable accuracies are 0.83 on IFEval and 0.54 on MATH Lvl 5. We then extend our approach to analyze task-dependent saturation and to probe contamination-related shifts on math reasoning tasks. Finally, we introduce a balanced I-optimal sampling algorithm that recovers near-full-data frontiers using roughly 20% of the parameter-count-weighted evaluation budget, as low as 5% on some tasks, while maintaining comparable calibration. Together, our work releases Proteus-2k, the latest model performance evaluation dataset, and introduces a practical methodology for translating compute budgets into reliable performance expectations and for monitoring when capability boundaries shift across time.
Read the full abstract

All authors Hanlin Zhang, Jikai Jin, Vasilis Syrgkanis, Sham Kakade

Computation and Language

Mnemis: Dual-Route Retrieval on Hierarchical Graphs for Long-Term LLM Memory

Zihao Tang, Xin Yu, Ziyu Xiao et al. (+9 more)

AI Memory, specifically how models organizes and retrieves historical messages, becomes increasingly valuable to Large Language Models (LLMs), yet existing methods (RAG and Graph-RAG) primarily retrieve memory through similarity-based mechanisms. While efficient, such System-1-style retrieval struggles with scenarios that require global reasoning or comprehensive coverage of all relevant information. In this work, We propose Mnemis, a novel memory framework that integrates System-1 similarity search with a complementary System-2 mechanism, termed Global Selection. Mnemis organizes memory into a base graph for similarity retrieval and a hierarchical graph that enables top-down, deliberate traversal over semantic hierarchies. By combining the complementary strength from both retrieval routes, Mnemis retrieves memory items that are both semantically and structurally relevant. Mnemis achieves state-of-the-art performance across all compared methods on long-term memory benchmarks, scoring 93.9 on LoCoMo and 91.6 on LongMemEval-S using GPT-4.1-mini.
Read the full abstract

All authors Zihao Tang, Xin Yu, Ziyu Xiao, Zengxuan Wen, Zelin Li, Jiaxi Zhou, Hualei Wang, Haohua Wang, Haizhen Huang, Weiwei Deng, Feng Sun, Qi Zhang

Computation and Language

Extracting Consumer Insight from Text: A Large Language Model Approach to Emotion and Evaluation Measurement

Stephan Ludwig, Peter J. Danaher, Xiaohao Yang et al. (+4 more)

Accurately measuring consumer emotions and evaluations from unstructured text remains a core challenge for marketing research and practice. This study introduces the Linguistic eXtractor (LX), a fine-tuned, large language model trained on consumer-authored text that also has been labeled with consumers' self-reported ratings of 16 consumption-related emotions and four evaluation constructs: trust, commitment, recommendation, and sentiment. LX consistently outperforms leading models, including GPT-4 Turbo, RoBERTa, and DeepSeek, achieving 81% macro-F1 accuracy on open-ended survey responses and greater than 95% accuracy on third-party-annotated Amazon and Yelp reviews. An application of LX to online retail data, using seemingly unrelated regression, affirms that review-expressed emotions predict product ratings, which in turn predict purchase behavior. Most emotional effects are mediated by product ratings, though some emotions, such as discontent and peacefulness, influence purchase directly, indicating that emotional tone provides meaningful signals beyond star ratings. To support its use, a no-code, cost-free, LX web application is available, enabling scalable analyses of consumer-authored text. In establishing a new methodological foundation for consumer perception measurement, this research demonstrates new methods for leveraging large language models to advance marketing research and practice, thereby achieving validated detection of marketing constructs from consumer data.
Read the full abstract

All authors Stephan Ludwig, Peter J. Danaher, Xiaohao Yang, Yu-Ting Lin, Ehsan Abedin, Dhruv Grewal, Lan Du

Computation and Language · Cross-listed from cs.LG

The Information Geometry of Softmax: Probing and Steering

Kiho Park, Todd Nief, Yo Joong Choe et al. (+1 more)

This paper concerns the question of how AI systems encode semantic structure into the geometric structure of their representation spaces. The motivating observation is that the natural geometry of these representation spaces should reflect the way models use representations to produce behavior. We focus on the important special case of representations that define softmax distributions. In this case, we argue that the natural geometry is information geometry. Our focus is on the role of information geometry on semantic encoding and the linear representation hypothesis. As an illustrative application, we develop "dual steering", a method for robustly steering representations to exhibit a particular concept using linear probes. We prove that dual steering optimally modifies the target concept while minimizing changes to off-target concepts. Empirically, we find that dual steering enhances the controllability and stability of concept manipulation.
Read the full abstract

All authors Kiho Park, Todd Nief, Yo Joong Choe, Victor Veitch

Computation and Language · Cross-listed from cs.CY

Stability Regimes for Framing-Sensitive Fine-Tuning in Language Models

Victor De Lima, Jiqun Liu and Grace Hui Yang

AI systems increasingly shape how individuals access and evaluate information, intensifying concerns about misinformation. Computational simulation offers a controlled complement to direct empirical study, with fine-tuned LLMs enabling powerful agent-based simulations. However, fine-tuning often induces broad behavioral shifts including bias drift, degenerate heuristics, and changes in baseline task behavior, which undermine experimental validity. The central challenge is inducing localized decision shifts while preserving the model's aggregate behavior outside the targeted intervention. We study this through a stability analysis of supervised fine-tuning, attenuating loss on selected training slices to shift false-claim acceptance. We introduce FrameRef, a large-scale dataset of semantically equivalent claims with controlled surface framings across five dimensions (Authoritative, Consensus, Emotional, Sensationalist, Prestige), with human validation confirming systematic framing effects on claim acceptance. We present a methodology for identifying model-specific stability regimes in which targeted error shifts can be induced while preserving aggregate accuracy and calibration, and demonstrate its application to selected language models. A sequential exposure task over FrameRef shows that small framing-conditioned shifts compound into substantially different cumulative outcomes under feedback, but largely disappear when feedback is removed, confirming that interventions modify specific decision tendencies rather than global performance. We release code, LoRA adapters, and data at https://github.com/infosenselab/framing-sensitive-ft.
Read the full abstract

Monday 16 February

Computation and Language

Fine-Tune, Don't Prompt, Your Language Model to Identify Biased Language in Clinical Notes

Isotta Landi, Eugenia Alleva, Nicole Bussola et al. (+5 more)

Clinical documentation can contain emotionally charged language with stigmatizing or privileging valences. We present a framework for detecting and classifying such language as stigmatizing, privileging, or neutral. We constructed a curated lexicon of biased terms scored for emotional valence. We then used lexicon-based matching to extract text chunks from OB-GYN delivery notes (Mount Sinai Hospital, NY) and MIMIC-IV discharge summaries across multiple specialties. Three clinicians annotated all chunks, enabling characterization of valence patterns across specialties and healthcare systems. We benchmarked multiple classification strategies (zero-shot prompting, in-context learning, and supervised fine-tuning) across encoder-only models (GatorTron) and generative large language models (Llama). Fine-tuning with lexically primed inputs consistently outperformed prompting approaches. GatorTron achieved an F1 score of 0.96 on the OB-GYN test set, outperforming larger generative models while requiring minimal prompt engineering and fewer computational resources. External validation on MIMIC-IV revealed limited cross-domain generalizability (F1 < 0.70, 44% drop). Training on the broader MIMIC-IV dataset improved generalizability when testing on OB-GYN (F1 = 0.71, 11% drop), but at the cost of reduced precision. Our findings demonstrate that fine-tuning outperforms prompting for emotional valence classification and that models must be adapted to specific medical specialties to achieve clinically appropriate performance. The same terms can carry different emotional valences across specialties: words with clinical meaning in one context may be stigmatizing in another. For bias detection, where misclassification risks undermining clinician trust or perpetuating patient harm, specialty-specific fine-tuning is essential to capture these semantic shifts. * Equal contribution.
Read the full abstract

All authors Isotta Landi, Eugenia Alleva, Nicole Bussola, Rebecca M. Cohen, Sarah Nowlin, Leslee J. Shaw, Alexander W. Charney, Kimberly B. Glazer

Computation and Language

Probing the Limits of the Lie Detector Approach to LLM Deception

Tom-Felix Thormann

A strand of mechanistic approaches to deception in large language models (LLMs) relies on "lie detectors" - truth probes aimed to identify model outputs that are internally represented as false. This lie detector approach to LLM deception implicitly assumes that deception is coextensive with lying. This paper challenges that assumption by experimentally investigating whether LLMs can produce deceptive statements that are not false and whether truth probes (if naively applied as deception detectors) fail to detect such behavior. In doing so, we make four core contributions: First, we provide a dataset of manually written examples of deceptive non-falsities grounded in linguistic theory. Second, we show across six open-source LLMs of different sizes and families that some models can select these deceptive non-falsities in a forced-choice task. Third, we demonstrate that while truth probes trained on a standard true-false dataset perform well at detecting false outputs, they perform significantly worse at detecting deceptive non-falsities. We view this as substantiating the worry that naive applications of truth probes as deception detectors suffer from a critical blind spot. Fourth, we explore whether training deception probes in dialogical contexts or including misleading non-false statements in the training improves the deception detection rate.
Read the full abstract

Computation and Language

SpreadsheetArena: Decomposing Preference in LLM Generation of Spreadsheet Workbooks

Srivatsa Kundurthy, Clara Na, Michael Handley et al. (+5 more)

We consider the task of end-to-end spreadsheet generation, where language models produce spreadsheet artifacts to satisfy users' explicit and implicit constraints, specified in natural language. We introduce SpreadsheetArena, a platform for evaluating models' performance on the task via blind pairwise preference votes of LLM-generated spreadsheet workbooks. As with other complex, open-ended tasks, relevant evaluation criteria can vary greatly across use cases, often in ways that are difficult to formalize. Compared to general dialogue or text generation settings, spreadsheet generation presents unique challenges and opportunities: the task output structure is well-defined and multi-dimensional, and there are often complex interactivity and layout considerations. We observe that stylistic, structural, and functional features of preferred spreadsheets vary meaningfully across prompts. Expert evaluations of spreadsheets for finance prompts suggest that even highly ranked models do not reliably produce spreadsheets aligned with domain-specific best practices. We host a live arena and release a dataset of prompts, generated spreadsheets, and preference votes, which we hope will facilitate further study of tasks operating over spreadsheets as a challenging and interesting class of complex, open-ended tasks for LLMs.
Read the full abstract

All authors Srivatsa Kundurthy, Clara Na, Michael Handley, Zach Kirshner, Chen Bo Calvin Zhang, Manasi Sharma, Emma Strubell, John Ling

Computation and Language

Leveraging Wikidata for Geographically Informed Sociocultural Bias Dataset Creation: Application to Latin America

Yannis Karmim, Renato Pino, Hernan Contreras et al. (+6 more)

Large Language Models (LLMs) exhibit inequalities with respect to various cultural contexts. Most prominent open-weights models are trained on Global North data and show prejudicial behavior towards other cultures. Moreover, there is a notable lack of resources to detect biases in non-English languages, especially from Latin America (Latam), a continent containing various cultures, even though they share a common cultural ground. We propose to leverage the content of Wikipedia, the structure of the Wikidata knowledge graph, and expert knowledge from social science in order to create a dataset of question/answer (Q/As) pairs, based on the different popular and social cultures of various Latin American countries. We create the LatamQA database of over 26k questions and associated answers extracted from 26k Wikipedia articles, and transformed into multiple-choice questions (MCQ) in Spanish and Portuguese, in turn translated to English. We use this MCQ to quantify the degree of knowledge of various LLMs and find out (i) a discrepancy in performances between the Latam countries, ones being easier than others for the majority of the models, (ii) that the models perform better in their original language, and (iii) that Iberian Spanish culture is better known than Latam one.
Read the full abstract

All authors Yannis Karmim, Renato Pino, Hernan Contreras, Hernan Lira, Sebastian Cifuentes, Simon Escoffier, Luis Martí, Djamé Seddah, Valentin Barrière

Computation and Language

Beyond the Prompt in Large Language Models: Comprehension, In-Context Learning, and Chain-of-Thought

Yuling Jiao, Yanming Lai, Huazhen Lin et al. (+3 more)

Large Language Models (LLMs) have demonstrated remarkable proficiency across diverse tasks, exhibiting emergent properties such as semantic prompt comprehension, In-Context Learning (ICL), and Chain-of-Thought (CoT) reasoning. Despite their empirical success, the theoretical mechanisms driving these phenomena remain poorly understood. This study dives into the foundations of these observations by addressing three critical questions: (1) How do LLMs accurately decode prompt semantics despite being trained solely on a next-token prediction objective? (2) Through what mechanism does ICL facilitate performance gains without explicit parameter updates? and (3) Why do intermediate reasoning steps in CoT prompting effectively unlock capabilities for complex, multi-step problems? Our results demonstrate that, through the autoregressive process, LLMs are capable of exactly inferring the transition probabilities between tokens across distinct tasks using provided prompts. We show that ICL enhances performance by reducing prompt ambiguity and facilitating posterior concentration on the intended task. Furthermore, we find that CoT prompting activates the model's capacity for task decomposition, breaking complex problems into a sequence of simpler sub-tasks that the model has mastered during the pretraining phase. By comparing their individual error bounds, we provide novel theoretical insights into the statistical superiority of advanced prompt engineering techniques.
Read the full abstract

All authors Yuling Jiao, Yanming Lai, Huazhen Lin, Wensen Ma, Houduo Qi, Defeng Sun

Computation and Language

A Retrieval-Augmented Language Assistant for Unmanned Aircraft Safety Assessment and Regulatory Compliance

Gabriele Immordino, Andrea Vaiuso and Marcello Righi

This paper presents the design and validation of a retrieval-based assistant that supports safety assessment, certification activities, and regulatory compliance for unmanned aircraft systems. The work is motivated by the growing complexity of drone operations and the increasing effort required by applicants and aviation authorities to apply established assessment frameworks, including the Specific Operations Risk Assessment and the Pre-defined Risk Assessment, in a consistent and efficient manner. The proposed approach uses a controlled text-based architecture that relies exclusively on authoritative regulatory sources. To enable traceable and auditable outputs, the assistant grounds each response in retrieved passages and enforces citation-driven generation. System-level controls address common failure modes of generative models, including fabricated statements, unsupported inferences, and unclear provenance, by separating evidence storage from language generation and by adopting conservative behavior when supporting documentation is insufficient. The assistant is intentionally limited to decision support; it does not replace expert judgment and it does not make autonomous determinations. Instead, it accelerates context-specific information retrieval and synthesis to improve document preparation and review while preserving human responsibility for critical conclusions. The architecture is implemented using established open-source components, and key choices in retrieval strategy, interaction constraints, and response policies are evaluated for suitability in safety-sensitive regulatory environments. The paper provides technical and operational guidance for integrating retrieval-based assistants into aviation oversight workflows while maintaining accountability, traceability, and regulatory compliance.
Read the full abstract

Computation and Language

Generating Realistic, Protocol-Compliant Maritime Radio Dialogues using Self-Instruct and Low-Rank Adaptation

Gürsel Akdeniz and Emin Cagatay Nakilcioglu

VHF radio miscommunication remains a major safety risk in maritime operations, with human factors accounting for over 58% of recorded incidents in Europe between 2014 and 2023. Despite decades of operational use, VHF radio communications are still prone to noise, interference, linguistic variability, and the absence of real-time transcription, making procedural errors both frequent and difficult to correct. Developing AI-assisted systems to support real-time communication and decision-making requires a considerable amount of high-quality maritime data, yet operational, regulatory, and privacy constraints render such datasets scarce. This study introduces a compliance aware Self-Instruct methodology for generating realistic maritime radio dialogues that conform to the IMO's SMCP. Our approach integrates a 26-filter verification pipeline directly into the iterative generation loop to enforce entity information accuracy, hallucination detection, SMCP-compliance, logical consistency, and linguistic diversity. We employ LORA for parameter-efficient fine-tuning, reducing computational overhead during training and enabling efficient deployment of the resulting models on resource-constrained maritime systems. To assess dataset quality, we introduce a novel evaluation framework combining automated and expert assessments: Format Accuracy, Information Accuracy, Uniqueness, and Logical Coherence. Experiments using publicly available vessel, coastal and AIS datasets demonstrate that the approach produces synthetically diverse, procedurally compliant, and operationally realistic dialogues. Although downstream applications such as automatic speech recognition and natural language processing are reserved for future work, the released code, datasets, and verification tools provide a reproducible foundation for artificial intelligence-assisted maritime safety and other safety-critical domains.
Read the full abstract

Computation and Language

Iterative LLM-based improvement for French Clinical Interview Transcription and Speaker Diarization

Ambre Marie, Thomas Bertin, Guillaume Dardenne et al. (+1 more)

Automatic speech recognition for French medical conversations remains challenging, with word error rates often exceeding 30% in spontaneous clinical speech. This study proposes a multi-pass LLM post-processing architecture alternating between Speaker Recognition and Word Recognition passes to improve transcription accuracy and speaker attribution. Ablation studies on two French clinical datasets (suicide prevention telephone counseling and preoperative awake neurosurgery consultations) investigate four design choices: model selection, prompting strategy, pass ordering, and iteration depth. Using Qwen3-Next-80B, Wilcoxon signed-rank tests confirm significant WDER reductions on suicide prevention conversations (p<0.05, n=18), while maintaining stability on awake neurosurgery consultations (n=10), with zero output failures and acceptable computational cost (RTF 0.32), suggesting feasibility for offline clinical deployment, pending validation on larger corpora.
Read the full abstract

All authors Ambre Marie, Thomas Bertin, Guillaume Dardenne, Gwenolé Quellec

Computation and Language

Task-Lens: Cross-Task Utility Based Speech Dataset Profiling for Low-Resource Indian Languages

Swati Sharma, Divya V. Sharma and Anubha Gupta

The rising demand for inclusive speech technologies amplifies the need for multilingual datasets for Natural Language Processing (NLP) research. However, limited awareness of existing task-specific resources in low-resource languages hinders research. This challenge is especially acute in linguistically diverse countries, such as India. Cross-task profiling of existing Indian speech datasets can alleviate the data scarcity challenge. This involves investigating the utility of datasets across multiple downstream tasks rather than focusing on a single task. Prior surveys typically catalogue datasets for a single task, leaving comprehensive cross-task profiling as an open opportunity. Therefore, we propose Task-Lens, a cross-task survey that assesses the readiness of 50 Indian speech datasets spanning 26 languages for nine downstream speech tasks. First, we analyze which datasets contain metadata and properties suitable for specific tasks. Next, we propose task-aligned enhancements to unlock datasets to their full downstream potential. Finally, we identify tasks and Indian languages that are critically underserved by current resources. Our findings reveal that many Indian speech datasets contain untapped metadata that can support multiple downstream tasks. By uncovering cross-task linkages and gaps, Task-Lens enables researchers to explore the broader applicability of existing datasets and to prioritize dataset creation for underserved tasks and languages.
Read the full abstract

Computation and Language · Cross-listed from cs.SD

Hello-Chat: Towards Realistic Social Audio Interactions

Yueran Hou, Peilei Jia, Zihan Sun et al. (+5 more)

Recent advancements in Large Audio Language Models (LALMs) have demonstrated exceptional performance in speech recognition and translation. However, existing models often suffer from a disconnect between perception and expression, resulting in a robotic "read-speech" style that lacks the spontaneity and emotional resonance of real human interaction. In this report, we introduce Hello-Chat, an end-to-end audio language model designed for realistic social scenarios. By leveraging a massive dataset of real-life conversations and employing a modality-interleaved training strategy, Hello-Chat achieves a breakthrough in anthropomorphic generation. Experimental results show that our model not only reaches state-of-the-art (SOTA) performance on specific audio understanding tasks but also significantly outperforms existing baselines in prosodic naturalness and emotional alignment, paving the way for the next generation of empathetic AI agents.
Read the full abstract

All authors Yueran Hou, Peilei Jia, Zihan Sun, Qihang Lu, Wenbing Yang, Yingming Gao, Ya Li, Jun Gao

Computation and Language · Cross-listed from cs.CV

How to Train Your Long-Context Visual Document Model

Austin Veselka

We present the first comprehensive, large-scale study of training long-context vision language models up to 344K context, targeting long-document visual question answering with measured transfer to long-context text. While several such strong are open-weight, namely Qwen3 VL and GLM 4.5/6V, their training recipes and data pipelines are not reproducible. We systematically study continued pretraining, supervised finetuning, and preference optimization for 24B and 32B parameter models, backed by extensive LC evaluations and ablations to bridge this gap, and achieve state-of-the-art performance on MMLongBenchDoc for both parameter scales. In addition to this, our key findings include: (i) training on context lengths that match evaluation context lengths outperforms training on longer contexts, (ii) training and evaluating with page indices provides a simple, high-impact boost to long-document performance, (iii) our synthetic data pipelines enable self-improvement via continued pretraining and supervised finetuning, and (iv) we extend the known text-to-visual long context transfer to the reverse, showing that visual long context training transfers to long-context text performance. We also release MMLBD-C, a manually corrected version of MMLongBenchDoc to reduce erroneous and low quality examples in the benchmark.
Read the full abstract

Computation and Language · Cross-listed from cs.MA

Colosseum: Auditing Collusion in Cooperative Multi-Agent Systems

Mason Nakamura, Abhinav Kumar, Saswat Das et al. (+5 more)

Multi-agent systems, where LLM agents communicate through free-form language, enable sophisticated coordination for solving complex cooperative tasks. This surfaces a unique safety problem when a group of agents forms a coalition and colludes to pursue secondary goals and degrade the joint objective. In this paper, we present Colosseum, a framework for auditing LLM agents' collusive behavior in multi-agent settings. We ground how agents cooperate through a formal multi-agent decision-making framework and measure action-based collusive behavior in actions via regret relative to the cooperative optimum and compare it with communication-based collusive behavior. Colosseum enables audits of LLM agents for collusion under benign settings, different coalition objectives, persuasion tactics, and network topologies. We then introduce a new behavioral probe by creating secret communication channels between agents, showing that most out-of-the-box models exhibit a propensity to collude under this probe, which we term emergent collusion. Furthermore, we discover “collusion on paper” when agents plan to collude in text but often pick non-collusive actions. Colosseum provides a new way to audit collusion in cooperative multi-agent systems while presenting observations about how collusion emerges, what affects collusion efficacy, and which strategies may mitigate it.
Read the full abstract

All authors Mason Nakamura, Abhinav Kumar, Saswat Das, Sahar Abdelnabi, Saaduddin Mahmud, Ferdinando Fioretto, Shlomo Zilberstein, Eugene Bagdasarian

Computation and Language

OpaqueToolsBench: Learning Nuances of Tool Behavior Through Interaction

Skyler Hallinan, Thejas Venkatesh, Xiang Ren et al. (+4 more)

Tool-calling is essential for Large Language Model (LLM) agents to complete real-world tasks. While most existing benchmarks assume simple, perfectly documented tools, real-world tools (e.g., general "search" APIs) are often opaque, lacking clear best practices or failure modes. Can LLM agents improve their performance in environments with opaque tools by interacting and subsequently improving documentation? To study this, we create OpaqueToolsBench, a benchmark consisting of three distinct task-oriented environments: general function calling, interactive chess playing, and long-trajectory agentic search. Each environment provides underspecified tools that models must learn to use effectively to complete the task. Results on OpaqueToolsBench suggest existing methods for automatically documenting tools are expensive and unreliable when tools are opaque. To address this, we propose a simple framework, ToolObserver, that iteratively refines tool documentation by observing execution feedback from tool-calling trajectories. Our approach outperforms existing methods on OpaqueToolsBench across datasets, even in relatively hard settings. Furthermore, for test-time tool exploration settings, our method is also efficient, consuming 3.5-7.5x fewer total tokens than the best baseline.
Read the full abstract

All authors Skyler Hallinan, Thejas Venkatesh, Xiang Ren, Sai Praneeth Karimireddy, Ashwin Paranjape, Yuhao Zhang, Jack Hessel

Computation and Language · Cross-listed from cs.CR

Weight space Detection of Backdoors in LoRA Adapters

David Puertolas Merenciano, Ekaterina Vasyagina, Kevin Zhu et al. (+2 more)

LoRA adapters let users fine-tune large language models (LLMs) efficiently. However, LoRA adapters are shared through open repositories like Hugging Face Hub \citep{huggingface_hub_docs}, making them vulnerable to backdoor attacks. Current detection methods require running the model with test input data – making them impractical for screening thousands of adapters where the trigger for backdoor behavior is unknown. We detect poisoned adapters by analyzing their weight matrices directly, without running the model – making our method trigger-agnostic. For each attention projection (Q, K, V, O), our method extracts five spectral statistics from the low-rank update $\Delta W$, yielding a 20-dimensional signature for each adapter. A logistic regression detector trained on this representation separates benign and poisoned adapters across three model families – Llama-3.2-3B~\citep{llama3}, Qwen2.5-3B~\citep{qwen25}, and Gemma-2-2B~\citep{gemma2} – on unseen test adapters drawn from instruction-following, reasoning, question-answering, code, and classification tasks. Across all three architectures, the detector achieves 100% accuracy.
Read the full abstract

All authors David Puertolas Merenciano, Ekaterina Vasyagina, Kevin Zhu, Javier Ferrando, Maheep Chaudhary

Computation and Language

AIC CTU@AVerImaTeC: dual-retriever RAG for image-text fact checking

Herbert Ullrich and Jan Drchal

In this paper, we present our 3rd place system in the AVerImaTeC shared task, which combines our last year's retrieval-augmented generation (RAG) pipeline with a reverse image search (RIS) module. Despite its simplicity, our system delivers competitive performance with a single multimodal LLM call per fact-check at just $0.013 on average using GPT5.1 via OpenAI Batch API. Our system is also easy to reproduce and tweak, consisting of only three decoupled modules - a textual retrieval module based on similarity search, an image retrieval module based on API-accessed RIS, and a generation module using GPT5.1 - which is why we suggest it as an accesible starting point for further experimentation. We publish its code and prompts, as well as our vector stores and insights into the scheme's running costs and directions for further improvement.
Read the full abstract

Computation and Language · Cross-listed from cs.IR

ScrapeGraphAI-100k: Dataset for Schema-Constrained LLM Generation

William Brach, Francesco Zuppichini, Marco Vinciguerra et al. (+1 more)

Producing output that conforms to a specified JSON schema underlies tool use, structured extraction, and knowledge base construction in modern large language models. Despite this centrality, public datasets for the task remain small, synthetic, or text-only, and rarely pair real page content with the prompts and schemas used in practice. We introduce ScrapeGraphAI-100k, 93,695 schema-constrained extraction events collected via opt-in ScrapeGraphAI telemetry in Q2–Q3 2025, deduplicated and balanced by schema from 9M raw events. The corpus spans 18 000+ unique schemas across 15 named languages plus a long-tail Other category, with English and Traditional Chinese covering 88% of detected content, each instance pairs Markdown-converted page content with a prompt, schema, LLM response, and per-example jsonschema-rs structural conformance labels (semantic correctness is out of scope, and raw HTML is deferred beyond v1.0). We characterize structural diversity across the corpus and identify sharp failure thresholds as schema complexity grows. As a case study, a 1.7B student fine-tuned on this data closely tracks the output distribution of its GPT-5-nano teacher, though it still trails a 30B-A3B reference (3.3B active parameters) on schema compliance. We offer this distillation result as preliminary evidence that grounding schema-constrained generation in real practitioner workloads at scale enables training and benchmarking that prior synthetic or text-only corpora could not support.
Read the full abstract

All authors William Brach, Francesco Zuppichini, Marco Vinciguerra, Lorenzo Padoan

Computation and Language · Cross-listed from cs.LG

Seeing to Generalize: How Visual Data Corrects Binding Shortcuts

Nicolas Buzeta, Felipe del Rio, Cristian Hinostroza et al. (+3 more)

Vision Language Models (VLMs) are designed to extend Large Language Models (LLMs) with visual capabilities, yet in this work we observe a surprising phenomenon: VLMs can outperform their underlying LLMs on purely text-only tasks, particularly in long-context information retrieval. To investigate this effect, we build a controlled synthetic retrieval task and find that a transformer trained only on text achieves perfect in-distribution accuracy but fails to generalize out of distribution, while subsequent training on an image-tokenized version of the same task nearly doubles text-only OOD performance. Mechanistic interpretability reveals that visual training changes the model's internal binding strategy: text-only training encourages positional shortcuts, whereas image-based training disrupts them through spatial translation invariance, forcing the model to adopt a more robust symbolic binding mechanism that persists even after text-only examples are reintroduced. We further characterize how binding strategies vary across training regimes, visual encoders, and initializations, and show that analogous shifts occur during pretrained LLM-to-VLM transitions. Our findings suggest that cross-modal training can enhance reasoning and generalization even for tasks grounded in a single modality.
Read the full abstract

All authors Nicolas Buzeta, Felipe del Rio, Cristian Hinostroza, Denis Parra, Hans Lobel, Rodrigo Toro Icarte