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

Thursday 5 February

Computation and Language

MedErrBench: A Fine-Grained Multilingual Benchmark for Medical Error Detection and Correction with Clinical Expert Annotations

Congbo Ma, Yichun Zhang, Yousef Al-Jazzazi et al. (+6 more)

Inaccuracies in existing or generated clinical text may lead to serious adverse consequences, especially if it is a misdiagnosis or incorrect treatment suggestion. With Large Language Models (LLMs) increasingly being used across diverse healthcare applications, comprehensive evaluation through dedicated benchmarks is crucial. However, such datasets remain scarce, especially across diverse languages and contexts. In this paper, we introduce MedErrBench, the first multilingual benchmark for error detection, localization, and correction, developed under the guidance of experienced clinicians. Based on an expanded taxonomy of ten common error types, MedErrBench covers English, Arabic and Chinese, with natural clinical cases annotated and reviewed by domain experts. We assessed the performance of a range of general-purpose, language-specific, and medical-domain language models across all three tasks. Our results reveal notable performance gaps, particularly in non-English settings, highlighting the need for clinically grounded, language-aware systems. By making MedErrBench and our evaluation protocols publicly-available, we aim to advance multilingual clinical NLP to promote safer and more equitable AI-based healthcare globally. The dataset is available in the supplementary material. An anonymized version of the dataset is available at: https://github.com/congboma/MedErrBench.
Read the full abstract

All authors Congbo Ma, Yichun Zhang, Yousef Al-Jazzazi, Ahamed Foisal, Laasya Sharma, Yousra Sadqi, Khaled Saleh, Jihad Mallat, Farah E. Shamout

Computation and Language

Modelling the Morphology of Verbal Paradigms: A Case Study in the Tokenization of Turkish and Hebrew

Giuseppe Samo and Paola Merlo

We investigate how transformer models represent complex verb paradigms in Turkish and Modern Hebrew, concentrating on how tokenization strategies shape this ability. Using the Blackbird Language Matrices task on natural data, we show that for Turkish – with its transparent morphological markers – both monolingual and multilingual models succeed, either when tokenization is atomic or when it breaks words into small subword units. For Hebrew, instead, monolingual and multilingual models diverge. A multilingual model using character-level tokenization fails to capture the language non-concatenative morphology, but a monolingual model with morpheme-aware segmentation performs well. Performance improves on more synthetic datasets, in all models.
Read the full abstract

Computation and Language · Cross-listed from cs.AI

Generative Ontology: When Structured Knowledge Learns to Create

Benny Cheung

Traditional ontologies describe domain structure but cannot generate novel artifacts. Large language models generate fluently but produce outputs lacking structural validity, hallucinating mechanisms without components, goals without end conditions. We introduce Generative Ontology, a framework synthesizing these complementary strengths: ontology provides the grammar; the LLM provides the creativity. Generative Ontology encodes domain knowledge as executable Pydantic schemas constraining LLM generation via DSPy signatures. A multi-agent pipeline assigns specialized roles: a Mechanics Architect designs game systems, a Theme Weaver integrates narrative, a Balance Critic identifies exploits, each carrying a professional "anxiety" that prevents shallow outputs. Retrieval-augmented generation grounds designs in precedents from existing exemplars. We demonstrate the framework through GameGrammar, generating complete tabletop game designs, and present three empirical studies. An ablation study (120 designs, 4 conditions) shows multi-agent specialization produces the largest quality gains (fun d=1.12, depth d=1.59; p<.001), while schema validation eliminates structural errors (d=4.78). A benchmark against 20 published board games reveals structural parity but a bounded creative gap (fun d=1.86): generated designs score 7-8 while published games score 8-9. A test-retest study (50 evaluations) validates the LLM-based evaluator, with 7/9 metrics achieving Good-to-Excellent reliability (ICC 0.836-0.989). The pattern generalizes beyond games. Any domain with expert vocabulary, validity constraints, and accumulated exemplars is a candidate for Generative Ontology.
Read the full abstract

Computation and Language

CASTLE: A Comprehensive Benchmark for Evaluating Student-Tailored Personalized Safety in Large Language Models

Rui Jia, Ruiyi Lan, Fengrui Liu et al. (+7 more)

Large language models (LLMs) have advanced the development of personalized learning in education. However, their inherent generation mechanisms often produce homogeneous responses to identical prompts. This one-size-fits-all mechanism overlooks the substantial heterogeneity in students cognitive and psychological, thereby posing potential safety risks to vulnerable groups. Existing safety evaluations primarily rely on context-independent metrics such as factual accuracy, bias, or toxicity, which fail to capture the divergent harms that the same response might cause across different student attributes. To address this gap, we propose the concept of Student-Tailored Personalized Safety and construct CASTLE based on educational theories. This benchmark covers 15 educational safety risks and 14 student attributes, comprising 92,908 bilingual scenarios. We further design three evaluation metrics: Risk Sensitivity, measuring the model ability to detect risks; Emotional Empathy, evaluating the model capacity to recognize student states; and Student Alignment, assessing the match between model responses and student attributes. Experiments on 18 SOTA LLMs demonstrate that CASTLE poses a significant challenge: all models scored below an average safety rating of 2.3 out of 5, indicating substantial deficiencies in personalized safety assurance.
Read the full abstract

All authors Rui Jia, Ruiyi Lan, Fengrui Liu, Zhongxiang Dai, Bo Jiang, Jing Shao, Jingyuan Chen, Guandong Xu, Fei Wu, Min Zhang

Computation and Language · Cross-listed from cs.LG

Rewards as Labels: Revisiting RLVR from a Classification Perspective

Zepeng Zhai, Meilin Chen, Jiaxuan Zhao et al. (+3 more)

Reinforcement Learning with Verifiable Rewards has recently advanced the capabilities of Large Language Models in complex reasoning tasks by providing explicit rule-based supervision. Among RLVR methods, GRPO and its variants have achieved strong empirical performance. Despite their success, we identify that they suffer from Gradient Misassignment in Positives and Gradient Domination in Negatives, which lead to inefficient and suboptimal policy updates. To address these issues, we propose Rewards as Labels (REAL), a novel framework that revisits verifiable rewards as categorical labels rather than scalar weights, thereby reformulating policy optimization as a classification problem. Building on this, we further introduce anchor logits to enhance policy learning. Our analysis reveals that REAL induces a monotonic and bounded gradient weighting, enabling balanced gradient allocation across rollouts and effectively mitigating the identified mismatches. Extensive experiments on mathematical reasoning benchmarks show that REAL improves training stability and consistently outperforms GRPO and strong variants such as DAPO. On the 1.5B model, REAL improves average Pass@1 over DAPO by 6.7%. These gains further scale to 7B model, REAL continues to outperform DAPO and GSPO by 6.2% and 1.7%, respectively. Notably, even with a vanilla binary cross-entropy, REAL remains stable and exceeds DAPO by 4.5% on average.
Read the full abstract

All authors Zepeng Zhai, Meilin Chen, Jiaxuan Zhao, Junlang Qian, Lei Shen, Yuan Lu

Computation and Language · Cross-listed from cs.HC

AI chatbots versus human healthcare professionals: a systematic review and meta-analysis of empathy in patient care

Alastair Howcroft, Amber Bennett-Weston, Ahmad Khan et al. (+3 more)

Background: Empathy is widely recognized for improving patient outcomes, including reduced pain and anxiety and improved satisfaction, and its absence can cause harm. Meanwhile, use of artificial intelligence (AI)-based chatbots in healthcare is rapidly expanding, with one in five general practitioners using generative AI to assist with tasks such as writing letters. Some studies suggest AI chatbots can outperform human healthcare professionals (HCPs) in empathy, though findings are mixed and lack synthesis. Sources of data: We searched multiple databases for studies comparing AI chatbots using large language models with human HCPs on empathy measures. We assessed risk of bias with ROBINS-I and synthesized findings using random-effects meta-analysis where feasible, whilst avoiding double counting. Areas of agreement: We identified 15 studies (2023-2024). Thirteen studies reported statistically significantly higher empathy ratings for AI, with only two studies situated in dermatology favouring human responses. Of the 15 studies, 13 provided extractable data and were suitable for pooling. Meta-analysis of those 13 studies, all utilising ChatGPT-3.5/4, showed a standardized mean difference of 0.87 (95% CI, 0.54-1.20) favouring AI (P < .00001), roughly equivalent to a two-point increase on a 10-point scale. Areas of controversy: Studies relied on text-based assessments that overlook non-verbal cues and evaluated empathy through proxy raters. Growing points: Our findings indicate that, in text-only scenarios, AI chatbots are frequently perceived as more empathic than human HCPs. Areas timely for developing research: Future research should validate these findings with direct patient evaluations and assess whether emerging voice-enabled AI systems can deliver similar empathic advantages.
Read the full abstract

All authors Alastair Howcroft, Amber Bennett-Weston, Ahmad Khan, Joseff Griffiths, Simon Gay, Jeremy Howick

Computation and Language · Cross-listed from cs.AI

BhashaSetu: Cross-Lingual Knowledge Transfer from High-Resource to Extreme Low-Resource Languages

Subhadip Maji and Arnab Bhattacharya

Despite remarkable advances in natural language processing, developing effective systems for low-resource languages remains a formidable challenge, with performances typically lagging far behind high-resource counterparts due to data scarcity and insufficient linguistic resources. Cross-lingual knowledge transfer has emerged as a promising approach to address this challenge by leveraging resources from high-resource languages. In this paper, we investigate methods for transferring linguistic knowledge from high-resource languages to low-resource languages, where the number of labeled training instances is in hundreds. We focus on sentence-level and word-level tasks. We introduce a novel method, GETR (Graph-Enhanced Token Representation) for cross-lingual knowledge transfer along with two adopted baselines (a) augmentation in hidden layers and (b) token embedding transfer through token translation. Experimental results demonstrate that our GNN-based approach significantly outperforms existing multilingual and cross-lingual baseline methods, achieving 13 percentage point improvements on truly low-resource languages (Mizo, Khasi) for POS tagging, and 20 and 27 percentage point improvements in macro-F1 on simulated low-resource languages (Marathi, Bangla, Malayalam) across sentiment classification and NER tasks respectively. We also present a detailed analysis of the transfer mechanisms and identify key factors that contribute to successful knowledge transfer in this linguistic context.
Read the full abstract

Computation and Language · Cross-listed from cs.SE

ArkTS-CodeSearch: A Open-Source ArkTS Dataset for Code Retrieval

Yulong He, Artem Ermakov, Sergey Kovalchuk et al. (+2 more)

ArkTS is a core programming language in the OpenHarmony ecosystem, yet research on ArkTS code intelligence is hindered by the lack of public datasets and evaluation benchmarks. This paper presents a large-scale ArkTS dataset constructed from open-source repositories, targeting code retrieval and code evaluation tasks. We design a single-search task, where natural language comments are used to retrieve corresponding ArkTS functions. ArkTS repositories are crawled from GitHub and Gitee, and comment-function pairs are extracted using tree-sitter-arkts, followed by cross-platform deduplication and statistical analysis of ArkTS function types. We further evaluate existing open-source code embedding models on the single-search task and perform fine-tuning using both ArkTS and TypeScript training datasets, resulting in a high-performing model for ArkTS code understanding. This work establishes the first systematic benchmark for ArkTS code retrieval. Both the dataset and our fine-tuned model are available at https://huggingface.co/hreyulog/embedinggemma_arkts and https://huggingface.co/datasets/hreyulog/arkts-code-docstring .
Read the full abstract

All authors Yulong He, Artem Ermakov, Sergey Kovalchuk, Artem Aliev, Dmitry Shalymov

Computation and Language

Multi-Task GRPO: Reliable LLM Reasoning Across Tasks

Shyam Sundhar Ramesh, Xiaotong Ji, Matthieu Zimmer et al. (+5 more)

RL-based post-training with GRPO is widely used to improve large language models on individual reasoning tasks. However, real-world deployment requires reliable performance across diverse tasks. A straightforward multi-task adaptation of GRPO often leads to imbalanced outcomes, with some tasks dominating optimization while others stagnate. Moreover, tasks can vary widely in how frequently prompts yield zero advantages (and thus zero gradients), which further distorts their effective contribution to the optimization signal. To address these issues, we propose a novel Multi-Task GRPO (MT-GRPO) algorithm that (i) dynamically adapts task weights to explicitly optimize worst-task performance and promote balanced progress across tasks, and (ii) introduces a ratio-preserving sampler to ensure task-wise policy gradients reflect the adapted weights. Experiments on both 3-task and 9-task settings show that MT-GRPO consistently outperforms baselines in worst-task accuracy. In particular, MT-GRPO achieves 16-28% and 6% absolute improvement on worst-task performance over standard GRPO and DAPO, respectively, while maintaining competitive average accuracy. Moreover, MT-GRPO requires 50% fewer training steps to reach 50% worst-task accuracy in the 3-task setting, demonstrating substantially improved efficiency in achieving reliable performance across tasks.
Read the full abstract

All authors Shyam Sundhar Ramesh, Xiaotong Ji, Matthieu Zimmer, Sangwoong Yoon, Zhiyong Wang, Haitham Bou Ammar, Aurelien Lucchi, Ilija Bogunovic

Computation and Language · Cross-listed from cs.LG

Steering Large Reasoning Models towards Concise Reasoning via Flow Matching

Yawei Li, Benjamin Bergner, Yinghan Zhao et al. (+3 more)

Large Reasoning Models (LRMs) excel at complex reasoning tasks, but their efficiency is often hampered by overly verbose outputs. Prior steering methods attempt to address this issue by applying a single, global vector to hidden representations – an approach grounded in the restrictive linear representation hypothesis. In this work, we introduce FlowSteer, a nonlinear steering method that goes beyond uniform linear shifts by learning a complete transformation between the distributions associated with verbose and concise reasoning. This transformation is learned via Flow Matching as a velocity field, enabling precise, input-dependent control over the model's reasoning process. By aligning steered representations with the distribution of concise-reasoning activations, FlowSteer yields more compact reasoning than the linear shifts. Across diverse reasoning benchmarks, FlowSteer demonstrates strong task performance and token efficiency compared to leading inference-time baselines. Our work demonstrates that modeling the full distributional transport with generative techniques offers a more effective and principled foundation for controlling LRMs.
Read the full abstract

All authors Yawei Li, Benjamin Bergner, Yinghan Zhao, Vihang Prakash Patil, Bei Chen, Cheng Wang

Computation and Language · Cross-listed from cs.LG

When Shared Knowledge Hurts: Spectral Over-Accumulation in Model Merging

Yayuan Li, Ze Peng, Jian Zhang et al. (+3 more)

Model merging combines multiple fine-tuned models into a single model by adding their weight updates, providing a lightweight alternative to retraining. Existing methods primarily target resolving conflicts between task updates, leaving the failure mode of over-counting shared knowledge unaddressed. We show that when tasks share aligned spectral directions (i.e., overlapping singular vectors), a simple linear combination repeatedly accumulates these directions, inflating the singular values and biasing the merged model toward shared subspaces. To mitigate this issue, we propose Singular Value Calibration (SVC), a training-free and data-free post-processing method that quantifies subspace overlap and rescales inflated singular values to restore a balanced spectrum. Across vision and language benchmarks, SVC consistently improves strong merging baselines and achieves state-of-the-art performance. Furthermore, by modifying only the singular values, SVC improves the performance of Task Arithmetic by 13.0%. Code is available at https://github.com/lyymuwu/SVC.
Read the full abstract

All authors Yayuan Li, Ze Peng, Jian Zhang, Jintao Guo, Yue Duan, Yinghuan Shi

Computation and Language · Cross-listed from cs.AI

A Unified Multimodal Framework for Dataset Construction and Model-Based Diagnosis of Ameloblastoma

Ajo Babu George, Anna Mariam John, Athul Anoop et al. (+1 more)

Artificial intelligence (AI)-enabled diagnostics in maxillofacial pathology require structured, high-quality multimodal datasets. However, existing resources provide limited ameloblastoma coverage and lack the format consistency needed for direct model training. We present a newly curated multimodal dataset specifically focused on ameloblastoma, integrating annotated radiological, histopathological, and intraoral clinical images with structured data derived from case reports. Natural language processing techniques were employed to extract clinically relevant features from textual reports, while image data underwent domain specific preprocessing and augmentation. Using this dataset, a multimodal deep learning model was developed to classify ameloblastoma variants, assess behavioral patterns such as recurrence risk, and support surgical planning. The model is designed to accept clinical inputs such as presenting complaint, age, and gender during deployment to enhance personalized inference. Quantitative evaluation demonstrated substantial improvements; variant classification accuracy increased from 46.2 percent to 65.9 percent, and abnormal tissue detection F1-score improved from 43.0 percent to 90.3 percent. Benchmarked against resources like MultiCaRe, this work advances patient-specific decision support by providing both a robust dataset and an adaptable multimodal AI framework.
Read the full abstract

All authors Ajo Babu George, Anna Mariam John, Athul Anoop, Balu Bhasuran

Computation and Language

A Human-in-the-Loop, LLM-Centered Architecture for Knowledge-Graph Question Answering

Larissa Pusch, Alexandre Courtiol and Tim Conrad

Large Language Models (LLMs) excel at language understanding but remain limited in knowledge-intensive domains due to hallucinations, outdated information, and limited explainability. Text-based retrieval-augmented generation (RAG) helps ground model outputs in external sources but struggles with multi-hop reasoning. Knowledge Graphs (KGs), in contrast, support precise, explainable querying, yet require a knowledge of query languages. This work introduces an interactive framework in which LLMs generate and explain Cypher graph queries and users iteratively refine them through natural language. Applied to real-world KGs, the framework improves accessibility to complex datasets while preserving factual accuracy and semantic rigor and provides insight into how model performance varies across domains. Our core quantitative evaluation is a 90-query benchmark on a synthetic movie KG that measures query explanation quality and fault detection across multiple LLMs, complemented by two smaller real-life query-generation experiments on a Hyena KG and the MaRDI (Mathematical Research Data Initiative) KG.
Read the full abstract

Computation and Language

Transport and Merge: Cross-Architecture Merging for Large Language Models

Chenhang Cui, Binyun Yang, Fei Shen et al. (+5 more)

Large language models (LLMs) achieve strong capabilities by scaling model capacity and training data, yet many real-world deployments rely on smaller models trained or adapted from low-resource data. This gap motivates the need for mechanisms to transfer knowledge from large, high-resource models to smaller, low-resource targets. While model merging provides an effective transfer mechanism, most existing approaches assume architecture-compatible models and therefore cannot directly transfer knowledge from large high-resource LLMs to heterogeneous low-resource targets. In this work, we propose a cross-architecture merging framework based on optimal transport (OT) that aligns activations to infer cross-neuron correspondences between heterogeneous models. The resulting transport plans are then used to guide direct weight-space fusion, enabling effective high-resource to low-resource transfer using only a small set of inputs. Extensive experiments across low-resource languages and specialized domains demonstrate consistent improvements over target models.
Read the full abstract

All authors Chenhang Cui, Binyun Yang, Fei Shen, Yuxin Chen, Jingnan Zheng, Xiang Wang, An Zhang, Tat-Seng Chua

Computation and Language

LinguistAgent Technical Report: A Reflective Multi-Model Platform for Automated Linguistic Annotation

Bingru Li

Data annotation remains a significant bottleneck in the field of humanities and social sciences, particularly for complex linguistic tasks such as metaphor identification. While Large Language Models (LLMs) show promise, a significant gap remains between the theoretical capability of LLMs and their practical utility for researchers. This paper introduces LinguistAgent, an integrated, user-friendly platform that leverages a reflective multi-model architecture to automate linguistic annotation. The platform comprises an Annotator and an optional Reviewer to simulate a peer-review process. This platform supports comparative experiments across three main paradigms: Prompt Engineering (Zero-shot/Few-shot/Chain-of-thought), Retrieval-Augmented Generation, and Fine-tuning. We demonstrate LinguistAgent's efficacy by replicating the task of metaphor identification from a published study, which provides real-time token-level evaluation (F1 and Cohen's kappa) against human gold standards. The application and codes are released on https://github.com/Bingru-Li/LinguistAgent.
Read the full abstract

Computation and Language

Reasoning under Ambiguity: Uncertainty-Aware Multilingual Emotion Classification under Partial Supervision

Md. Mithun Hossain, Mashary N. Alrasheedy, Nirban Bhowmick et al. (+4 more)

Contemporary knowledge-based systems increasingly rely on multilingual emotion identification to support intelligent decision-making, yet they face major challenges due to emotional ambiguity and incomplete supervision. Emotion recognition from text is inherently uncertain because multiple emotional states often co-occur and emotion annotations are frequently missing or heterogeneous. Most existing multi-label emotion classification methods assume fully observed labels and rely on deterministic learning objectives, which can lead to biased learning and unreliable predictions under partial supervision. This paper introduces Reasoning under Ambiguity, an uncertainty-aware framework for multilingual multi-label emotion classification that explicitly aligns learning with annotation uncertainty. The proposed approach uses a shared multilingual encoder with language-specific optimization and an entropy-based ambiguity weighting mechanism that down-weights highly ambiguous training instances rather than treating missing labels as negative evidence. A mask-aware objective with positive-unlabeled regularization is further incorporated to enable robust learning under partial supervision. Experiments on English, Spanish, and Arabic emotion classification benchmarks demonstrate consistent improvements over strong baselines across multiple evaluation metrics, along with improved training stability, robustness to annotation sparsity, and enhanced interpretability.
Read the full abstract

All authors Md. Mithun Hossain, Mashary N. Alrasheedy, Nirban Bhowmick, Shamim Forhad, Md. Shakil Hossain, Sudipto Chaki, Md Shafiqul Islam

Computation and Language · Cross-listed from cs.CV

MerNav: A Highly Generalizable Memory-Execute-Review Framework for Zero-Shot Object Goal Navigation

Dekang Qi, Shuang Zeng, Xinyuan Chang et al. (+4 more)

Visual Language Navigation (VLN) is one of the fundamental capabilities for embodied intelligence and a critical challenge that urgently needs to be addressed. However, existing methods are still unsatisfactory in terms of both success rate (SR) and generalization: Supervised Fine-Tuning (SFT) approaches typically achieve higher SR, while Training-Free (TF) approaches often generalize better, but it is difficult to obtain both simultaneously. To this end, we propose a Memory-Execute-Review framework. It consists of three parts: a hierarchical memory module for providing information support, an execute module for routine decision-making and actions, and a review module for handling abnormal situations and correcting behavior. We validated the effectiveness of this framework on the Object Goal Navigation task. Across 4 datasets, our average SR achieved absolute improvements of 7% and 5% compared to all baseline methods under TF and Zero-Shot (ZS) settings, respectively. On the most commonly used HM3D_v0.1 and the more challenging open vocabulary dataset HM3D_OVON, the SR improved by 8% and 6%, under ZS settings. Furthermore, on the MP3D and HM3D_OVON datasets, our method not only outperformed all TF methods but also surpassed all SFT methods, achieving comprehensive leadership in both SR (5% and 2%) and generalization. Additionally, we deployed the MerNav model on the humanoid robot and conducted experiments in the real world. The project address is: https://qidekang.github.io/MerNav.github.io/
Read the full abstract

All authors Dekang Qi, Shuang Zeng, Xinyuan Chang, Feng Xiong, Shichao Xie, Xiaolong Wu, Mu Xu

Computation and Language

Structured Context Engineering for File-Native Agentic Systems: Evaluating Schema Accuracy, Format Effectiveness, and Multi-File Navigation at Scale

Damon McMillan

Large Language Model agents increasingly operate external systems through programmatic interfaces, yet practitioners lack empirical guidance on how to structure the context these agents consume. Using SQL generation as a proxy for programmatic agent operations, we present a systematic study of context engineering for structured data, comprising 9,649 experiments across 11 models, 4 formats (YAML, Markdown, JSON, Token-Oriented Object Notation [TOON]), and schemas ranging from 10 to 10,000 tables. Our findings challenge common assumptions. First, architecture choice is model-dependent: file-based context retrieval improves accuracy for frontier-tier models (Claude, GPT, Gemini; +2.7%, p=0.029) but shows mixed results for open source models (aggregate -7.7%, p<0.001), with deficits varying substantially by model. Second, format does not significantly affect aggregate accuracy (chi-squared=2.45, p=0.484), though individual models, particularly open source, exhibit format-specific sensitivities. Third, model capability is the dominant factor, with a 21 percentage point accuracy gap between frontier and open source tiers that dwarfs any format or architecture effect. Fourth, file-native agents scale to 10,000 tables through domain-partitioned schemas while maintaining high navigation accuracy. Fifth, file size does not predict runtime efficiency: compact or novel formats can incur a token overhead driven by grep output density and pattern unfamiliarity, with the magnitude depending on model capability. These findings provide practitioners with evidence-based guidance for deploying LLM agents on structured systems, demonstrating that architectural decisions should be tailored to model capability rather than assuming universal best practices.
Read the full abstract

Computation and Language

Causal Front-Door Adjustment for Robust Jailbreak Attacks on LLMs

Yao Zhou, Zeen Song, Wenwen Qiang et al. (+2 more)

Jailbreak attacks on safety-aligned LLMs are increasingly effective, yet attack success alone does not reveal whether compliance is driven by preserved task intent, suppressed safety signals, or attack-specific artifacts. We use Pearl's front-door adjustment to motivate separating task-relevant semantics from safety-related variation. We propose Causal Front-Door Adjustment Attack (CFA^2), a white-box framework that operationalizes this perspective in LLM latent space using Sparse Autoencoders, contrastive datasets, and weight orthogonalization. After offline calibration, CFA^2 enables efficient generation without per-prompt iterative optimization, performs strongly on HarmBench across four open-source LLMs, and provides source-side signals for same-family black-box prompt transfer.
Read the full abstract

All authors Yao Zhou, Zeen Song, Wenwen Qiang, Fengge Wu, Changwen Zheng

Computation and Language

Once Correct, Still Wrong: Counterfactual Hallucination in Multilingual Vision-Language Models

Basel Mousi, Fahim Dalvi, Shammur Chowdhury et al. (+2 more)

Vision-language models (VLMs) can achieve high accuracy while still accepting culturally plausible but visually incorrect interpretations. Existing hallucination benchmarks rarely test this failure mode, particularly outside Western contexts and English. We introduce M$^2$CQA, a culturally grounded multimodal benchmark built from images spanning 17 MENA countries, paired with contrastive true and counterfactual statements in English, Arabic, and its dialects. To isolate hallucination beyond raw accuracy, we propose the CounterFactual Hallucination Rate (CFHR), which measures counterfactual acceptance conditioned on correctly answering the true statement. Evaluating state-of-the-art VLMs under multiple prompting strategies, we find that CFHR rises sharply in Arabic, especially in dialects, even when true-statement accuracy remains high. Moreover, reasoning-first prompting consistently increases counterfactual hallucination, while answering before justifying improves robustness. We make the dataset publicly available for the community (https://huggingface.co/datasets/QCRI/M2CQA)).
Read the full abstract

All authors Basel Mousi, Fahim Dalvi, Shammur Chowdhury, Firoj Alam, Nadir Durrani

Computation and Language

Grammatical Error Correction Evaluation by Optimally Transporting Edit Representation

Takumi Goto, Yusuke Sakai and Taro Watanabe

Automatic evaluation in grammatical error correction (GEC) is crucial for selecting the best-performing systems. Currently, reference-based metrics are a popular choice, which basically measure the similarity between hypothesis and reference sentences. However, similarity measures based on embeddings, such as BERTScore, are often ineffective, since many words in the source sentences remain unchanged in both the hypothesis and the reference. This study focuses on edits specifically designed for GEC, i.e., ERRANT, and computes similarity measured over the edits from the source sentence. To this end, we propose edit vector, a representation for an edit, and introduce a new metric, UOT-ERRANT, which transports these edit vectors from hypothesis to reference using unbalanced optimal transport. Experiments with SEEDA meta-evaluation show that UOT-ERRANT improves evaluation performance, particularly in the +Fluency domain where many edits occur. Moreover, our method is highly interpretable because the transport plan can be interpreted as a soft edit alignment, making UOT-ERRANT a useful metric for both system ranking and analyzing GEC systems. Our code is available from https://github.com/gotutiyan/uot-errant.
Read the full abstract

Computation and Language · Cross-listed from cs.IR

SciDef: Datasets and Tools for Automated Definition Extraction from Scientific Literature with LLMs

Filip Kučera, Christoph Mandl, Isao Echizen et al. (+2 more)

Scientific concepts are often defined inconsistently across papers, making it difficult to compare findings, reuse terminology, and build reliable downstream resources. We present SciDef, a resource suite for scientific definition extraction. The suite contains DefExtra, a benchmark of 268 human-validated author-stated definitions from 75 academic papers; DefSim, 60 human-labeled definition-pair similarity judgments; and an open LLM-based pipeline for PDF preprocessing, chunking, definition extraction, prompt optimization, and evaluation. We validate the resources by benchmarking 16 language models across prompting strategies and chunking schemes. The strongest set-level configuration achieves a score of 0.397, while the highest-coverage configuration matches at least one prediction to 86.4% of gold definitions but over-generates candidate definitions. We further show that an NLI-based matching metric agrees strongly with human DefSim judgments. These results position SciDef as a reusable benchmark and tooling layer for definition-centric literature analysis, while highlighting relevance-aware filtering as the key bottleneck for fully automatic definition extraction. Code & datasets are available at https://github.com/Media-Bias-Group/SciDef.
Read the full abstract

All authors Filip Kučera, Christoph Mandl, Isao Echizen, Radu Timofte, Timo Spinde

Computation and Language · Cross-listed from cs.AI

H-AdminSim: A Multi-Agent Simulator for Realistic Hospital Administrative Workflows with FHIR Integration

Jun-Min Lee, Meong Hi Son and Edward Choi

Hospital administration departments handle a wide range of operational tasks and, in large hospitals, process over 10,000 requests per day, driving growing interest in LLM-based automation. However, prior work has focused primarily on patient-physician interactions or isolated administrative subtasks, failing to capture the complexity of real administrative workflows. To address this gap, we propose H-AdminSim, a comprehensive simulation framework that combines realistic data generation with multi-agent-based simulation of hospital administrative workflows. These tasks are quantitatively evaluated using detailed rubrics, enabling systematic comparison of LLMs. Through FHIR integration, H-AdminSim provides a unified and interoperable environment for testing administrative workflows across heterogeneous hospital settings, serving as a standardized testbed for assessing the feasibility and performance of LLM-driven administrative automation.
Read the full abstract

Computation and Language

OPUS: Towards Efficient and Principled Data Selection in Large Language Model Pre-training in Every Iteration

Shaobo Wang, Xuan Ouyang, Tianyi Xu et al. (+9 more)

As high-quality public text approaches exhaustion, a phenomenon known as the Data Wall, pre-training is shifting from more tokens to better tokens. However, existing methods either rely on heuristic static filters that ignore training dynamics, or use dynamic yet optimizer-agnostic criteria based on raw gradients. We propose OPUS (Optimizer-induced Projected Utility Selection), a dynamic data selection framework that defines utility in the optimizer-induced update space. OPUS scores candidates by projecting their effective updates, shaped by modern optimizers, onto a target direction derived from a stable, in-distribution proxy. To ensure scalability, we employ Ghost technique with CountSketch for computational efficiency, and Boltzmann sampling for data diversity, incurring only 4.7% additional compute overhead. OPUS achieves remarkable results across diverse corpora, quality tiers, optimizers, and model scales. In pre-training of GPT-2 Large/XL on FineWeb and FineWeb-Edu with 30B tokens, OPUS outperforms industrial-level baselines and even full 200B-token training. Moreover, when combined with industrial-level static filters, OPUS further improves pre-training efficiency, even with lower-quality data. Furthermore, in continued pre-training of Qwen3-8B-Base on SciencePedia, OPUS achieves superior performance using only 0.5B tokens compared to full training with 3B tokens, demonstrating significant data efficiency gains in specialized domains.
Read the full abstract

All authors Shaobo Wang, Xuan Ouyang, Tianyi Xu, Yuzheng Hu, Jialin Liu, Guo Chen, Tianyu Zhang, Junhao Zheng, Kexin Yang, Xingzhang Ren, Dayiheng Liu, Linfeng Zhang

Computation and Language

Late-to-Early Training: LET LLMs Learn Earlier, So Faster and Better

Ji Zhao, Yufei Gu, Shitong Shao et al. (+3 more)

As Large Language Models (LLMs) achieve remarkable empirical success through scaling model and data size, pretraining has become increasingly critical yet computationally prohibitive, hindering rapid development. Despite the availability of numerous pretrained LLMs developed at significant computational expense, a fundamental real-world question remains underexplored: Can we leverage existing small pretrained models to accelerate the training of larger models? In this paper, we propose a Late-to-Early Training (LET) paradigm that enables LLMs to explicitly learn later knowledge in earlier steps and earlier layers. The core idea is to guide the early layers of an LLM during early training using representations from the late layers of a pretrained (i.e. late training phase) model. We identify two key mechanisms that drive LET's effectiveness: late-to-early-step learning and late-to-early-layer learning. These mechanisms significantly accelerate training convergence while robustly enhancing both language modeling capabilities and downstream task performance, enabling faster training with superior performance. Extensive experiments on 1.4B and 7B parameter models demonstrate LET's efficiency and effectiveness. Notably, when training a 1.4B LLM on the Pile dataset, our method achieves up to 1.6$\times$ speedup with nearly 5% improvement in downstream task accuracy compared to standard training, even when using a pretrained model with 10$\times$ fewer parameters than the target model.
Read the full abstract

All authors Ji Zhao, Yufei Gu, Shitong Shao, Xun Zhou, Liang Xiang, Zeke Xie

Computation and Language

Beyond Length: Context-Aware Expansion and Independence as Developmentally Sensitive Evaluation in Child Utterances

Jiyun Chun, Eric Fosler-Lussier, Michael White et al. (+1 more)

Evaluating the quality of children's utterances in adult-child dialogue remains challenging due to insufficient context-sensitive metrics. Common proxies such as Mean Length of Utterance (MLU), lexical diversity (vocd-D), and readability indices (Flesch-Kincaid Grade Level, Gunning Fog Index) are dominated by length and ignore conversational context, missing aspects of response quality such as reasoning depth, topic maintenance, and discourse planning. We introduce an LLM-as-a-judge framework that first classifies the Previous Adult Utterance Type and then scores the child's response along two axes: Expansion (contextual elaboration and inferential depth) and Independence (the child's contribution to advancing the discourse). These axes reflect fundamental dimensions in child language development, where Expansion captures elaboration, clause combining, and causal and contrastive connectives. Independence captures initiative, topic control, decreasing reliance on adult scaffolding through growing self-regulation, and audience design. We establish developmental validity by showing age-related patterns and demonstrate predictive value by improving age estimation over common baselines. We further confirm semantic sensitivity by detecting differences tied to discourse relations. Our metrics align with human judgments, enabling large-scale evaluation. This shifts child utterance assessment from simply measuring length to evaluating how meaningfully the child's speech contributes to and advances the conversation within its context.
Read the full abstract

All authors Jiyun Chun, Eric Fosler-Lussier, Michael White, Andrew Perrault

Computation and Language

IESR:Efficient MCTS-Based Modular Reasoning for Text-to-SQL with Large Language Models

Tao Liu, Jiafan Lu, Bohan Yu et al. (+11 more)

Text-to-SQL is a key natural language processing task that maps natural language questions to SQL queries, enabling intuitive interaction with web-based databases. Although current methods perform well on benchmarks like BIRD and Spider, they struggle with complex reasoning, domain knowledge, and hypothetical queries, and remain costly in enterprise deployment. To address these issues, we propose a framework named IESR(Information Enhanced Structured Reasoning) for lightweight large language models: (i) leverages LLMs for key information understanding and schema linking, and decoupling mathematical computation and SQL generation, (ii) integrates a multi-path reasoning mechanism based on Monte Carlo Tree Search (MCTS) with majority voting, and (iii) introduces a trajectory consistency verification module with a discriminator model to ensure accuracy and consistency. Experimental results demonstrate that IESR achieves state-of-the-art performance on the complex reasoning benchmark LogicCat (24.28 EX) and the Archer dataset (37.28 EX) using only compact lightweight models without fine-tuning. Furthermore, our analysis reveals that current coder models exhibit notable biases and deficiencies in physical knowledge, mathematical computation, and common-sense reasoning, highlighting important directions for future research. We released code at https://github.com/Ffunkytao/IESR-SLM.
Read the full abstract

All authors Tao Liu, Jiafan Lu, Bohan Yu, Pengcheng Wu, Liu Haixin, Guoyu Xu, Li Xiangheng, Lixiao Li, Jiaming Hou, Zhao Shijun, Xinglin Lyu, Kunli Zhang, Yuxiang Jia, Hongyin Zan

Computation and Language

Cross-Lingual Empirical Evaluation of Large Language Models for Arabic Medical Tasks

Chaimae Abouzahir, Congbo Ma, Nizar Habash et al. (+1 more)

In recent years, Large Language Models (LLMs) have become widely used in medical applications, such as clinical decision support, medical education, and medical question answering. Yet, these models are often English-centric, limiting their robustness and reliability for linguistically diverse communities. Recent work has highlighted discrepancies in performance in low-resource languages for various medical tasks, but the underlying causes remain poorly understood. In this study, we conduct a cross-lingual empirical analysis of LLM performance on Arabic and English medical question and answering. Our findings reveal a persistent language-driven performance gap that intensifies with increasing task complexity. Tokenization analysis exposes structural fragmentation in Arabic medical text, while reliability analysis suggests that model-reported confidence and explanations exhibit limited correlation with correctness. Together, these findings underscore the need for language-aware design and evaluation strategies in LLMs for medical tasks.
Read the full abstract

All authors Chaimae Abouzahir, Congbo Ma, Nizar Habash, Farah E. Shamout

Computation and Language

Mining or Synthesis? Rethinking Exploration Efficiency in Iterative Alignment of Mathematical Reasoning

Jun Rao, Zixiong Yu, Xuebo Liu et al. (+6 more)

Iterative Direct Preference Optimization (DPO) has emerged as a widely used paradigm for aligning Large Language Models on reasoning tasks. Existing approaches typically rely on Best-of-N sampling ($N\geq8$) to mine positive trajectories from the distribution tail. In this work, we show that in mathematical reasoning, increasing $N$ yields diminishing returns while increasing verifier-induced false-positive risk and the distribution shift required for policy updates. To address this, we introduce PACE (Proximal Alignment via Corrective Exploration), a generation-based corrective framework that replaces exhaustive mining with low-budget exploration ($2\leq N\leq3$). Rather than searching for increasingly rare positive samples, PACE synthesizes high-fidelity preference pairs from failed explorations through corrective hindsight refinement and verification-guided filtering. Empirically, PACE matches or exceeds the performance of DPO-R1 ($N=16$) while using about $1/5$ of the compute, and remains robust under 20% label corruption, where high-$N$ baselines exhibit substantially higher noise exploitation.
Read the full abstract

All authors Jun Rao, Zixiong Yu, Xuebo Liu, Guhan Chen, Jing Li, Hejin Wang, Jiansheng Wei, Xiaojun Meng, Min Zhang

Computation and Language · Cross-listed from cs.IR

Multi-Field Tool Retrieval

Yichen Tang, Weihang Su, Yiqun Liu et al. (+1 more)

Integrating external tools enables Large Language Models (LLMs) to interact with real-world environments and solve complex tasks. Given the growing scale of available tools, effective tool retrieval is essential to mitigate constraints of LLMs' context windows and ensure computational efficiency. Existing approaches typically treat tool retrieval as a traditional ad-hoc retrieval task, matching user queries against the entire raw tool documentation. In this paper, we identify three fundamental challenges that limit the effectiveness of this paradigm: (i) the incompleteness and structural inconsistency of tool documentation; (ii) the significant semantic and granular mismatch between user queries and technical tool documents; and, most importantly, (iii) the multi-aspect nature of tool utility, that involves distinct dimensions, such as functionality, input constraints, and output formats, varying in format and importance. To address these challenges, we introduce Multi-Field Tool Retrieval, a framework designed to align user intent with tool representations through fine-grained, multi-field modeling. Experimental results show that our framework achieves SOTA performance on five datasets and a mixed benchmark, exhibiting superior generalizability and robustness.
Read the full abstract

All authors Yichen Tang, Weihang Su, Yiqun Liu, Qingyao Ai