Peer Revenue Fully funded

Papers

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

Computer Science Computation and Language

cs.CL Watch Computation and Language

Thursday 27 November

Computation and Language · Cross-listed from cs.CV

From Compound Figures to Medical Multi-image Reasoning: Scaling Multimodal Large Language Models with Biomedical Literature

Zhen Chen, Yihang Fu, Rong Zhou et al. (+12 more)

Multimodal large language models (MLLMs) are increasingly capable in medical imaging, yet most focus on single-image settings. Clinical interpretation often requires integrating evidence across multiple images, such as different modalities, views, or time points. However, large-scale medical multi-image data and training strategies for such reasoning remain limited. We construct PMC-MI, a large-scale resource comprising 234,956 instruction instances derived from biomedical compound figures, with 10,555 multi-subimage instances structured for reinforcement learning and assessed by medical reviewers. We also introduce PMC-MI-Bench, a manually reviewed benchmark separated at the source-article level. We further propose a three-stage training framework, instantiated as M3LLM, combining supervised fine-tuning on multi-image instructions, selection-aware reinforcement learning for question-conditioned visual-evidence selection, and supervised consolidation on a broader instruction mixture. Evaluating M3LLM on PMC-MI-Bench, two public medical benchmarks, longitudinal chest radiographs, and a retrospective dermatology cohort reveals it achieves the highest performance among representative MLLMs. On PMC-MI-Bench, M3LLM scored 81.6, 84.6, and 85.2 on single-subimage, multi-subimage, and relative-position tasks, respectively, outperforming the strongest baselines (75.6, 79.1, and 77.6). It also set new records on the provenance-screened OmniMedVQA (88.6%) and MMMU-Med (68.0%), alongside top performance in two clinical settings after task-specific adaptation. These findings support using literature-derived multi-image supervision and targeted post-training for medical multi-image MLLMs. Data and code are publicly available.
Read the full abstract

All authors Zhen Chen, Yihang Fu, Rong Zhou, Serina Applebaum, Min Kyu Kim, Aidan Gilson, Morten Lee, Salahudeen Mirza, Gabriel Madera, Mauro Giuffre, Yuanting Pan, Roy Jiang, Hyunjae Kim, Hua Xu, Qingyu Chen

Computation and Language

Focused Chain-of-Thought: Efficient LLM Reasoning via Structured Input Information

Lukas Struppek, Dominik Hintersdorf, Hannah Struppek et al. (+2 more)

Recent large language models achieve strong reasoning performance by generating detailed chain-of-thought traces, but this often leads to excessive token use and high inference latency. Existing efficiency approaches typically focus on model-centric interventions, such as reinforcement learning or supervised fine-tuning, to reduce verbosity. In contrast, we propose a training-free, input-centric approach. Inspired by cognitive psychology, we introduce Focused Chain-of-Thought (F-CoT), which separates information extraction from the reasoning process. F-CoT first organizes the essential information from a query into a concise, structured context and then guides the model to reason exclusively over this context. By preventing attention to irrelevant details, F-CoT naturally produces shorter reasoning paths. On arithmetic word problems, F-CoT reduces generated tokens by 2-3x while maintaining accuracy comparable to standard zero-shot CoT. These results highlight structured input as a simple yet effective lever for more efficient LLM reasoning.
Read the full abstract

All authors Lukas Struppek, Dominik Hintersdorf, Hannah Struppek, Daniel Neider, Kristian Kersting

Computation and Language

RefineBench: Evaluating Refinement Capability of Language Models via Checklists

Young-Jun Lee, Seungone Kim, Byung-Kwan Lee et al. (+6 more)

Can language models (LMs) self-refine their own responses? This question is increasingly relevant as a wide range of real-world user interactions involve refinement requests. However, prior studies have largely tested LMs' refinement abilities on verifiable tasks such as competition math or symbolic reasoning with simplified scaffolds, whereas users often pose open-ended queries and provide varying degrees of feedback on what they desire. The recent advent of reasoning models that exhibit self-reflection patterns in their chains-of-thought further motivates this question. To analyze this, we introduce RefineBench, a benchmark of 1,000 challenging problems across 11 domains paired with a checklist-based evaluation framework. We evaluate two refinement modes: (1) guided refinement, where an LM is provided natural language feedback, and (2) self-refinement, where LMs attempt to improve without guidance. In the self-refinement setting, even frontier LMs such as Gemini 2.5 Pro and GPT-5 achieve modest baseline scores of 31.3% and 29.1%, respectively, and most models fail to consistently improve across iterations (e.g., Gemini-2.5-Pro gains only +1.8%, while DeepSeek-R1 declines by -0.1%). By contrast, in guided refinement, both proprietary LMs and large open-weight LMs (>70B) can leverage targeted feedback to refine responses to near-perfect levels within five turns. These findings suggest that frontier LMs require breakthroughs to self-refine their incorrect responses, and that RefineBench provides a valuable testbed for tracking progress.
Read the full abstract

All authors Young-Jun Lee, Seungone Kim, Byung-Kwan Lee, Minkyeong Moon, Yechan Hwang, Jong Myoung Kim, Graham Neubig, Sean Welleck, Ho-Jin Choi

Computation and Language

Lips-Jaw and Tongue-Jaw Articulatory Tradeoff in DYNARTmo

Bernd J. Kröger

This paper investigates how the dynamic articulatory model DYNARTmo accounts for articulatory tradeoffs between primary and secondary articulators, with a focus on lips-jaw and tongue-jaw coordination. While DYNARTmo does not implement full task-dynamic second-order biomechanics, it adopts first-order task-space gesture specifications comparable to those used in articulatory phonology and integrates a simplified mechanism for distributing articulatory effort across multiple articulators. We first outline the conceptual relationship between task dynamics and DYNARTmo, emphasizing the distinction between high-level task-space trajectories and their low-level articulatory execution. We then present simulation results for a set of CV syllables that illustrate how jaw displacement varies as a function of both place of articulation (labial, apical, dorsal) and vowel context (/a/, /i/, /u/). The model reproduces empirically attested patterns of articulatory synergy, including jaw-supported apical closures, lower-lip elevation in bilabial stops, tongue-jaw co-movement, and saturation effects in labial constrictions. These results demonstrate that even with computationally simplified assumptions, DYNARTmo can generate realistic spatio-temporal movement patterns that capture key aspects of articulatory tradeoff and synergy across a range of consonant-vowel combinations.
Read the full abstract

Computation and Language

Simplex-Optimized Hybrid Ensemble for Large Language Model Text Detection Under Generative Distribution Drif

Sepyan Purnama Kristanto, Lutfi Hakim and Dianni Yusuf

The widespread adoption of large language models (LLMs) has made it difficult to distinguish human writing from machine-produced text in many real applications. Detectors that were effective for one generation of models tend to degrade when newer models or modified decoding strategies are introduced. In this work, we study this lack of stability and propose a hybrid ensemble that is explicitly designed to cope with changing generator distributions. The ensemble combines three complementary components: a RoBERTa-based classifier fine-tuned for supervised detection, a curvature-inspired score based on perturbing the input and measuring changes in model likelihood, and a compact stylometric model built on hand-crafted linguistic features. The outputs of these components are fused on the probability simplex, and the weights are chosen via validation-based search. We frame this approach in terms of variance reduction and risk under mixtures of generators, and show that the simplex constraint provides a simple way to trade off the strengths and weaknesses of each branch. Experiments on a 30000 document corpus drawn from several LLM families including models unseen during training and paraphrased attack variants show that the proposed method achieves 94.2% accuracy and an AUC of 0.978. The ensemble also lowers false positives on scientific articles compared to strong baselines, which is critical in educational and research settings where wrongly flagging human work is costly
Read the full abstract

Computation and Language · Cross-listed from cs.LG

From Topology to Retrieval: Decoding Embedding Spaces with Unified Signatures

Florian Rottach, William Rudman, Bastian Rieck et al. (+2 more)

Studying the shape of text embedding spaces enhances our understanding of model behavior and how geometry affects downstream task performance. While a range of topological and geometric metrics have been explored, they are typically analyzed in isolation, and the connections between them remain largely unknown. In this paper, we present Unified Topological Signatures (UTS) for characterizing the structure of embedding spaces. We evaluate our method across a wide range of text embedding models and retrieval datasets, finding that it outperforms individual metrics in downstream performance prediction. Through several experiments, we demonstrate that the signatures characterize embedding spaces more accurately, improve the prediction of model-specific properties and biases, and reveal novel insights into representational similarity that challenge the global claims of the Platonic Representation Hypothesis.
Read the full abstract

All authors Florian Rottach, William Rudman, Bastian Rieck, Harrisen Scells, Carsten Eickhoff

Computation and Language

C$^2$DLM: Causal Concept-Guided Diffusion Large Language Models

Kairong Han, Nuanqiao Shan, Ziyu Zhao et al. (+6 more)

Autoregressive (AR) language models and Diffusion Language Models (DLMs) constitute the two principal paradigms of large language models. However, both paradigms suffer from insufficient reasoning capabilities. Human reasoning inherently relies on causal knowledge and thought, which are reflected in natural language. But in the AR paradigm, language is modeled as next token prediction (a strictly left-to-right, token-by-token order), whereas natural language itself exhibits more flexible causal structures. In the DLM paradigm, the attention mechanism is fully connected, which entirely disregards causal order. To fill this gap, we propose a \underline{C}ausal \underline{C}oncept-Guided \underline{D}iffusion \underline{L}anguage \underline{M}odel (C$^2$DLM). Starting from DLM's fully connected attention, C$^2$DLM first obtains a concept-level causal graph from the teacher model, and then explicitly guides attention to learn causal relationships between concepts. By focusing on causal relationships and avoiding interference from difficult subgoals involving causal inversion, C$^2$DLM improves 12% with about 3.2 times training speedup in the COT-OrderPerturb task, and achieves an average gain of 1.31% across six downstream reasoning tasks. More details in the repository ~\href{https://github.com/Kairong-Han/C-2-DLM}{here}.
Read the full abstract

All authors Kairong Han, Nuanqiao Shan, Ziyu Zhao, Zijing Hu, Xinpeng Dong, Junjian Ye, Lujia Pan, Fei Wu, Kun Kuang

Computation and Language

Bridging the Modality Gap by Similarity Standardization with Pseudo-Positive Samples

Shuhei Yamashita, Daiki Shirafuji and Tatsuhiko Saito

Advances in vision-language models (VLMs) have enabled effective cross-modality retrieval. However, when both text and images exist in the database, similarity scores would differ in scale by modality. This phenomenon, known as the modality gap, hinders accurate retrieval. Most existing studies address this issue with manually labeled data, e.g., by fine-tuning VLMs on them. In this work, we propose a similarity standardization approach with pseudo data construction. We first compute the mean and variance of the similarity scores between each query and its paired data in text or image modality. Using these modality-specific statistics, we standardize all similarity scores to compare on a common scale across modalities. These statistics are calculated from pseudo pairs, which are constructed by retrieving the text and image candidates with the highest cosine similarity to each query. We evaluate our method across seven VLMs using two multi-modal QA benchmarks (MMQA and WebQA), where each question requires retrieving either text or image data. Our experimental results show that our method significantly improves retrieval performance, achieving average Recall@20 gains of 64% on MMQA and 28% on WebQA when the query and the target data belong to different modalities. Compared to E5-V, which addresses the modality gap through image captioning, we confirm that our method more effectively bridges the modality gap.
Read the full abstract

Computation and Language

A Hybrid Theory and Data-driven Approach to Persuasion Detection with Large Language Models

Gia Bao Hoang, Keith J Ransom, Rachel Stephens et al. (+3 more)

Traditional psychological models of belief revision focus on face-to-face interactions, but with the rise of social media, more effective models are needed to capture belief revision at scale, in this rich text-based online discourse. Here, we use a hybrid approach, utilizing large language models (LLMs) to develop a model that predicts successful persuasion using features derived from psychological experiments. Our approach leverages LLM generated ratings of features previously examined in the literature to build a random forest classification model that predicts whether a message will result in belief change. Of the eight features tested, epistemic emotion and willingness to share were the top-ranking predictors of belief change in the model. Our findings provide insights into the characteristics of persuasive messages and demonstrate how LLMs can enhance models of successful persuasion based on psychological theory. Given these insights, this work has broader applications in fields such as online influence detection and misinformation mitigation, as well as measuring the effectiveness of online narratives.
Read the full abstract

All authors Gia Bao Hoang, Keith J Ransom, Rachel Stephens, Carolyn Semmler, Nicolas Fay, Lewis Mitchell

Computation and Language

Early Risk Prediction with Temporally and Contextually Grounded Clinical Language Processing

Rochana Chaturvedi, Yue Zhou, Andrew D. Boyd et al. (+5 more)

Clinical notes in Electronic Health Records (EHRs) capture rich temporal information on events, clinician reasoning, and lifestyle factors often missing from structured data. Leveraging them for predictive modeling can be impactful for timely identification of chronic diseases. However, they present core natural language processing (NLP) challenges: long text, irregular event distribution, complex temporal dependencies, privacy constraints, and resource limitations. We present two complementary methods for temporally and contextually grounded risk prediction from longitudinal notes. First, we introduce HiTGNN, a hierarchical temporal graph neural network that integrates intra-note temporal event structures, inter-visit dynamics, and medical knowledge to model patient trajectories with fine-grained temporal granularity. Second, we propose ReVeAL, a lightweight test-time framework that distills LLMs' reasoning into smaller verifier models. Applied to opportunistic screening for Type 2 Diabetes (T2D) using temporally realistic cohorts curated from private and public hospital corpora, HiTGNN achieves the highest predictive accuracy, especially for near-term risk, while preserving privacy and limiting reliance on large proprietary models. ReVeAL enhances sensitivity to true T2D cases and retains explanatory reasoning. Our ablations confirm the value of temporal structure and knowledge augmentation, and fairness analysis shows HiTGNN performs more equitably across subgroups.
Read the full abstract

All authors Rochana Chaturvedi, Yue Zhou, Andrew D. Boyd, Brian T. Layden, Mudassir Rashid, Lu Cheng, Ali Cinar, Barbara Di Eugenio

Computation and Language

ResearchArcade: Graph Interface for Academic Tasks

Jingjun Xu, Chongshan Lin, Haofei Yu et al. (+2 more)

Academic research generates diverse data sources, and as researchers increasingly use machine learning to assist research tasks, a crucial question arises: Can we build a unified data interface to support the development of machine learning models for various academic tasks? Models trained on such a unified interface can better support human researchers throughout the research process, eventually accelerating knowledge discovery. In this work, we introduce ResearchArcade, a graph-based interface that connects multiple academic data sources, unifies task definitions, and supports a wide range of base models to address key academic challenges. ResearchArcade utilizes a coherent multi-table format with graph structures to organize data from different sources, including academic corpora from ArXiv and peer reviews from OpenReview, while capturing information with multiple modalities, such as text, figures, and tables. ResearchArcade also preserves temporal evolution at both the manuscript and community levels, supporting the study of paper revisions as well as broader research trends over time. Additionally, ResearchArcade unifies diverse academic task definitions and supports various models with distinct input requirements. Our experiments across six academic tasks demonstrate that combining cross-source and multi-modal information enables a broader range of tasks, while incorporating graph structures consistently improves performance over baseline methods. This highlights the effectiveness of ResearchArcade and its potential to advance research progress.
Read the full abstract

All authors Jingjun Xu, Chongshan Lin, Haofei Yu, Tao Feng, Jiaxuan You

Computation and Language

AfriStereo: A Culturally Grounded Dataset for Evaluating Stereotypical Bias in Large Language Models

Yann Le Beux, Oluchi Audu, Oche D. Ankeli et al. (+4 more)

Existing AI bias evaluation benchmarks largely reflect Western perspectives, leaving African contexts underrepresented and enabling harmful stereotypes in applications across various domains. To address this gap, we introduce AfriStereo, the first open-source African stereotype dataset and evaluation framework grounded in local socio-cultural contexts. Through community engaged efforts across Senegal, Kenya, and Nigeria, we collected 1,163 stereotypes spanning gender, ethnicity, religion, age, and profession. Using few-shot prompting with human-in-the-loop validation, we augmented the dataset to over 5,000 stereotype-antistereotype pairs. Entries were validated through semantic clustering and manual annotation by culturally informed reviewers. Preliminary evaluation of language models reveals that nine of eleven models exhibit statistically significant bias, with Bias Preference Ratios (BPR) ranging from 0.63 to 0.78 (p <= 0.05), indicating systematic preferences for stereotypes over antistereotypes, particularly across age, profession, and gender dimensions. Domain-specific models appeared to show weaker bias in our setup, suggesting task-specific training may mitigate some associations. Looking ahead, AfriStereo opens pathways for future research on culturally grounded bias evaluation and mitigation, offering key methodologies for the AI community on building more equitable, context-aware, and globally inclusive NLP technologies.
Read the full abstract

All authors Yann Le Beux, Oluchi Audu, Oche D. Ankeli, Dhananjay Balakrishnan, Melissah Weya, Marie D. Ralaiarinosy, Ignatius Ezeani

Wednesday 26 November

Computation and Language

CAT: A Metric-Driven Framework for Analyzing the Consistency-Accuracy Relation of LLMs under Controlled Input Variations

Paulo Cavalin, Cassia Sanctos, Marcelo Grave et al. (+2 more)

We introduce CAT, a framework designed to evaluate and visualize the interplay of accuracy and response consistency of Large Language Models (LLMs) under controllable input variations, using multiple-choice (MC) benchmarks as a case study. Current evaluation practices primarily focus on model capabilities such as accuracy or benchmark scores and, more recently, measuring consistency is being considered an essential property for deploying LLMs in high-stake, real-world applications. We argue in this paper that although both dimensions should still be evaluated independently, their inter-dependency also need to be considered for a more nuanced evaluation of LLMs. At the core of CAT are the Consistency-Accuracy Relation (CAR) curves, which visualize how model accuracy varies with increasing consistency requirements, as defined by the Minimum-Consistency Accuracy (MCA) metric. We further propose the Consistency-Oriented Robustness Estimate (CORE) index, a global metric that combines the area and shape of the CAR curve to quantify the trade-off between accuracy and consistency. We present a practical demonstration of our framework across a diverse set of generalist and domain-specific LLMs, evaluated on multiple MC benchmarks. We also outline how CAT can be extended beyond MC tasks to support long-form, open-ended evaluations through adaptable scoring functions.
Read the full abstract

All authors Paulo Cavalin, Cassia Sanctos, Marcelo Grave, Claudio Pinhanez, Yago Primerano

Computation and Language · Cross-listed from cs.AI

MegaRAG: Multimodal Knowledge Graph-Based Retrieval Augmented Generation

Chi-Hsiang Hsiao, Yi-Cheng Wang, Tzung-Sheng Lin et al. (+2 more)

Retrieval-augmented generation (RAG) enables large language models (LLMs) to dynamically access external information, which is powerful for answering questions over previously unseen documents. Nonetheless, they struggle with high-level conceptual understanding and holistic comprehension due to limited context windows, which constrain their ability to perform deep reasoning over long-form, domain-specific content such as full-length books. To solve this problem, knowledge graphs (KGs) have been leveraged to provide entity-centric structure and hierarchical summaries, offering more structured support for reasoning. However, existing KG-based RAG solutions remain restricted to text-only inputs and fail to leverage the complementary insights provided by other modalities such as vision. On the other hand, reasoning from visual documents requires textual, visual, and spatial cues into structured, hierarchical concepts. To address this issue, we introduce a multimodal knowledge graph-based RAG that enables cross-modal reasoning for better content understanding. Our method incorporates visual cues into the construction of knowledge graphs, the retrieval phase, and the answer generation process. Experimental results across both global and fine-grained question answering tasks show that our approach consistently outperforms existing RAG-based approaches on both textual and multimodal corpora.
Read the full abstract

All authors Chi-Hsiang Hsiao, Yi-Cheng Wang, Tzung-Sheng Lin, Yi-Ren Yeh, Chu-Song Chen

Computation and Language

Graph-O1 : Monte Carlo Tree Search with Reinforcement Learning for Text-Attributed Graph Reasoning

Lihui Liu

ChatGPT said: Text-attributed graphs, where nodes and edges contain rich textual information, are widely used across diverse domains. A central challenge in this setting is question answering, which requires jointly leveraging unstructured text and the structured relational signals within the graph. Although Large Language Models (LLMs) have made significant advances in natural language understanding, their direct use for reasoning over text-attributed graphs remains limited. Retrieval-augmented generation methods that operate purely on text often treat passages as isolated units, ignoring the interconnected structure of the graph. Conversely, graph-based RAG methods that serialize large subgraphs into long textual sequences quickly become infeasible due to LLM context-length constraints, resulting in fragmented reasoning and degraded accuracy. To overcome these limitations, we introduce Graph-O1, an agentic GraphRAG framework that enables LLMs to conduct stepwise, interactive reasoning over graphs. Our approach integrates Monte Carlo Tree Search (MCTS) with end-to-end reinforcement learning, allowing the model to selectively explore and retrieve only the most informative subgraph components. The reasoning procedure is framed as a multi-turn interaction between the agent and the graph environment, and the agent is trained through a unified reward mechanism. Extensive experiments across multiple LLM backbones demonstrate that Graph-O1 consistently surpasses state-of-the-art baselines, producing answers that are more accurate, reliable, and interpretable.
Read the full abstract

Computation and Language

Towards Reasoning-Preserving Unlearning in Multimodal Large Language Models

Hongji Li, Junchi yao, Manjiang Yu et al. (+4 more)

Machine unlearning aims to erase requested data from trained models without full retraining. For Reasoning Multimodal Large Language Models (RMLLMs), this is uniquely challenging: intermediate chain-of-thought steps can still leak sensitive information even when final answers are forgotten, and overly aggressive interventions easily damage general reasoning ability. Yet no benchmark jointly evaluates how well unlearning methods suppress reasoning-level leakage while preserving reasoning competence. We address this gap with RMLLMU-Bench, the first benchmark for RMLLM unlearning that extends standard forgetting metrics with dedicated measures of reasoning leakage and reasoning retention. A systematic evaluation on RMLLMU-Bench reveals that existing unlearning methods for MLLMs and Large (Language) Reasoning Models (LRMs) either leave substantial leakage in the reasoning process or severely degrade reasoning performance. To address these gaps, we propose R-MUSE (Reasoning-preserving MLLM Unlearning via Subspace guidance and Adaptive Steering), a training-free and inference-time intervention framework that steers internal representations to forget both answers and reasoning traces while explicitly preserving general reasoning. Experiments on RMLLMU-Bench demonstrate that R-MUSE achieves a substantially better balance between effective forgetting and reasoning retention.
Read the full abstract

All authors Hongji Li, Junchi yao, Manjiang Yu, Priyanka Singh, Xue Li, Di Wang, Lijie Hu

Computation and Language

Reinforcement Learning for Latent-Space Thinking in LLMs

Enes Özeren and Matthias Aßenmacher

Chain-of-Thought (CoT) reasoning typically utilizes the discrete language space for thinking, which is inherently inefficient, as many generated tokens only enforce linguistic rules that are not required for reasoning. To bypass this, latent-space thinking allows models to think using the continuous embedding space. While existing methods for training those models show domain-specific gains, they fail to maintain performance in complex tasks, such as mathematical reasoning. We experimentally demonstrate that the Coconut approach, a form of supervised fine-tuning for latent-space thinking, is highly sensitive to design choices and exhibits several inherent limitations. To address these issues, we investigate reinforcement learning (RL) techniques – an underexplored direction in latent-space thinking – including GRPO and design a novel Latent RL method for directly optimizing the latent thinking steps. Our experimental results reveal that these RL-trained models still lag behind traditional language-space CoT models in the mathematical reasoning domain. We make our codebase publicly available.
Read the full abstract

Computation and Language · Cross-listed from cs.CY

AI Urban Scientist: Multi-Agent Collaborative Automation for Urban Research

Tong Xia, Jiankun Zhang, Ruiwen You et al. (+8 more)

Urban research aims to understand how cities operate and evolve as complex adaptive systems. With the rapid growth of urban data and analytical methodologies, the central challenge of the field has shifted from data availability to the integration of heterogeneous data into coherent, verifiable urban knowledge through multidisciplinary approaches. Recent advances in AI, particularly the emergence of large language models (LLMs), have enabled the development of AI scientists capable of autonomous reasoning, hypothesis generation, and data-driven experimentation, demonstrating substantial potential for autonomous urban research. However, most general-purpose AI systems remain misaligned with the domain-specific knowledge, methodological conventions, and inferential standards required in urban studies. Here, we introduce the AI Urban Scientist, a knowledge-driven multi-agent framework designed to support autonomous urban research. Grounded in hypotheses, peer-review feedback, datasets, and research methodologies distilled from large-scale prior studies, the system constructs structured domain knowledge that guides LLM-based agents to automatically generate hypotheses, identify and integrate multi-source urban datasets, conduct empirical analyses and simulations, and iteratively refine analytical methods. Through this process, the framework synthesizes new insights in urban science and accelerates the urban research lifecycle.
Read the full abstract

All authors Tong Xia, Jiankun Zhang, Ruiwen You, Ao Xu, Linghao Zhang, Tengyao Tu, Jingzhi Wang, Jinghua Piao, Yunke Zhang, Fengli Xu, Yong Li

Computation and Language

Beyond Confidence: Adaptive and Coherent Decoding for Diffusion Language Models

Kecheng Chen, Ziru Liu, Xijia Tao et al. (+7 more)

Diffusion Language Models (DLMs) have recently achieved significant success due to their any-order generation capabilities. However, existing inference methods typically rely on local, immediate-step metrics such as confidence or entropy which inherently lack a more reliable perspective. This limitation frequently leads to inconsistent sampling trajectories and suboptimal generation quality. To address this, we propose Coherent Contextual Decoding (CCD), a novel inference framework built upon two core innovations. First, CCD employs a trajectory rectification mechanism that leverages historical context to enhance sequence coherence, enabling the early rejection of suboptimal paths. We demonstrate that this mechanism is theoretically equivalent to modeling the consistency of historical steps via the conditional mutual information between context and token predictions. Building on this theoretical insight, we further address the inefficiency of conventional uniform decoding budgets. Instead of rigid allocations based on diffusion steps, we introduce an adaptive sampling strategy that dynamically adjusts the unmasking budget for each step according to our consistency metric. Consequently, our method significantly improves the quality of generation trajectories while accelerating the sampling process. Empirically, our method achieves a simultaneous enhancement in both inference speed and performance across diverse benchmarks on Dream and LLaDA, delivering up to 3.48x speedup alongside 3.91% performance improvement.
Read the full abstract

All authors Kecheng Chen, Ziru Liu, Xijia Tao, Hui Liu, Xinyu Fu, Suiyun Zhang, Dandan Tu, Lingpeng Kong, Rui Liu, Haoliang Li

Computation and Language

Start Making Sense(s): A Developmental Probe of Attention Specialization Using Lexical Ambiguity

Pamela D. Rivière and Sean Trott

Despite an in-principle understanding of self-attention matrix operations in Transformer language models (LMs), it remains unclear precisely how these operations map onto interpretable computations or functions–and how or when individual attention heads develop specialized attention patterns. Here, we present a pipeline to systematically probe attention mechanisms, and we illustrate its value by leveraging lexical ambiguity–where a single word has multiple meanings–to isolate attention mechanisms that contribute to word sense disambiguation. We take a "developmental" approach: first, using publicly available Pythia LM checkpoints, we identify inflection points in disambiguation performance for each LM in the suite; in 14M and 410M, we identify heads whose attention to disambiguating words covaries with overall disambiguation performance across development. We then stress-test the robustness of these heads to stimulus perturbations: in 14M, we find limited robustness, but in 410M, we identify multiple heads with surprisingly generalizable behavior. Then, in a causal analysis, we find that ablating the target heads demonstrably impairs disambiguation performance, particularly in 14M. We additionally reproduce developmental analyses of 14M across all of its random seeds. Together, these results suggest: that disambiguation benefits from a constellation of mechanisms, some of which (especially in 14M) are highly sensitive to the position and part-of-speech of the disambiguating cue; and that larger models (410M) may contain heads with more robust disambiguation behavior. They also join a growing body of work that highlights the value of adopting a developmental perspective when probing LM mechanisms.
Read the full abstract

Computation and Language

A Comparative Study of LLM Prompting and Fine-Tuning for Cross-genre Authorship Attribution on Chinese Lyrics

Yuxin Li, Lorraine Xu and Meng Fan Wang

We propose a novel study on authorship attribution for Chinese lyrics, a domain where clean, public datasets are sorely lacking. Our contributions are twofold: (1) we create a new, balanced dataset of Chinese lyrics spanning multiple genres, and (2) we develop and fine-tune a domain-specific model, comparing its performance against zero-shot inference using the DeepSeek LLM. We test two central hypotheses. First, we hypothesize that a fine-tuned model will outperform a zero-shot LLM baseline. Second, we hypothesize that performance is genre-dependent. Our experiments strongly confirm Hypothesis 2: structured genres (e.g. Folklore & Tradition) yield significantly higher attribution accuracy than more abstract genres (e.g. Love & Romance). Hypothesis 1 receives only partial support: fine-tuning improves robustness and generalization in Test1 (real-world data and difficult genres), but offers limited or ambiguous gains in Test2, a smaller, synthetically-augmented set. We show that the design limitations of Test2 (e.g., label imbalance, shallow lexical differences, and narrow genre sampling) can obscure the true effectiveness of fine-tuning. Our work establishes the first benchmark for cross-genre Chinese lyric attribution, highlights the importance of genre-sensitive evaluation, and provides a public dataset and analytical framework for future research. We conclude with recommendations: enlarge and diversify test sets, reduce reliance on token-level data augmentation, balance author representation across genres, and investigate domain-adaptive pretraining as a pathway for improved attribution performance.
Read the full abstract

Computation and Language

Tracing How Annotators Think: Augmenting Preference Judgments with Reading Processes

Karin de Langis, William Walker, Khanh Chi Le et al. (+1 more)

We propose an annotation approach that captures not only labels but also the reading process underlying annotators' decisions, e.g., what parts of the text they focus on, re-read or skim. Using this framework, we conduct a case study on the preference annotation task, creating a dataset PreferRead that contains fine-grained annotator reading behaviors obtained from mouse tracking. PreferRead enables detailed analysis of how annotators navigate between a prompt and two candidate responses before selecting their preference. We find that annotators re-read a response in roughly half of all trials, most often revisiting the option they ultimately choose, and rarely revisit the prompt. Reading behaviors are also significantly related to annotation outcomes: re-reading is associated with higher inter-annotator agreement, whereas long reading paths and times are associated with lower agreement. These results demonstrate that reading processes provide a complementary cognitive dimension for understanding annotator reliability, decision-making and disagreement in complex, subjective NLP tasks. Our code and data are publicly available.
Read the full abstract

All authors Karin de Langis, William Walker, Khanh Chi Le, Dongyeop Kang

Computation and Language

A Customer Journey in the Land of Oz: Leveraging the Wizard of Oz Technique to Model Emotions in Customer Service Interactions

Sofie Labat, Thomas Demeester and Véronique Hoste

Emotion-aware customer service needs in-domain conversational data, rich annotations, and predictive capabilities, but existing resources for emotion recognition are often out-of-domain, narrowly labeled, and focused on post-hoc detection. To address this, we conducted a controlled Wizard of Oz (WOZ) experiment to elicit interactions with targeted affective trajectories. The resulting corpus, EmoWOZ-CS, contains 2,148 bilingual (Dutch-English) written dialogues from 179 participants across commercial aviation, e-commerce, online travel agencies, and telecommunication scenarios. Our contributions are threefold: (1) Evaluate WOZ-based operator-steered valence trajectories as a design for emotion research; (2) Quantify human annotation performance and variation, including divergences between self-reports and third-party judgments; (3) Benchmark detection and forward-looking emotion inference in real-time support. Findings show neutral dominates participant messages; desire and gratitude are the most frequent non-neutral emotions. Agreement is moderate for multilabel emotions and valence, lower for arousal and dominance; self-reports diverge notably from third-party labels, aligning most for neutral, gratitude, and anger. Objective strategies often elicit neutrality or gratitude, while suboptimal strategies increase anger, annoyance, disappointment, desire, and confusion. Some affective strategies (cheerfulness, gratitude) foster positive reciprocity, whereas others (apology, empathy) can also leave desire, anger, or annoyance. Temporal analysis confirms successful conversation-level steering toward prescribed trajectories, most distinctly for negative targets; positive and neutral targets yield similar final valence distributions. Benchmarks highlight the difficulty of forward-looking emotion inference from prior turns, underscoring the complexity of proactive emotion-aware support.
Read the full abstract

Computation and Language

Improving Score Reliability of Multiple Choice Benchmarks with Consistency Evaluation and Altered Answer Choices

Paulo Cavalin, Cassia Sanctos, Marcelo Grave et al. (+2 more)

In this work we present the Consistency-Rebalanced Accuracy (CoRA) metric, improving the reliability of Large Language Model (LLM) scores computed on multiple choice (MC) benchmarks. Our metric explores the response consistency of the LLMs, taking advantage of synthetically-generated questions with altered answer choices. With two intermediate scores, i.e. Bare-Minimum-Consistency Accuracy (BMCA) and Consistency Index (CI), CoRA is computed by adjusting the multiple-choice question answering (MCQA) scores to better reflect the level of consistency of the LLM. We present evaluations in different benchmarks using diverse LLMs, and not only demonstrate that LLMs can present low response consistency even when they present high MCQA scores, but also that CoRA can successfully scale down the scores of inconsistent models.
Read the full abstract

All authors Paulo Cavalin, Cassia Sanctos, Marcelo Grave, Claudio Pinhanez, Yago Primerano

Computation and Language

FLAWS: A Benchmark for Error Identification and Localization in Scientific Papers

Sarina Xi, Vishisht Rao, Justin Payan et al. (+1 more)

The identification and localization of errors is a core task in peer review, yet the exponential growth of scientific output has made it increasingly difficult for human reviewers to reliably detect errors given the limited pool of experts. Recent advances in Large Language Models (LLMs) have sparked interest in their potential to support such evaluation tasks, from academic peer review to automated scientific assessment. However, despite the growing use of LLMs in review systems, their capabilities to pinpoint errors remain underexplored. In this work, we introduce Fault Localization Across Writing in Science (FLAWS), an automated benchmark consisting of 713 paper-error pairs designed to evaluate how effectively LLMs detect errors that undermine key claims in research papers. We construct the benchmark by systematically inserting claim-invalidating errors into peer-reviewed papers using LLMs, paired with an automated evaluation metric that measures whether models can identify and localize these errors. Developing such a benchmark presents unique challenges that we overcome: ensuring that the inserted errors are well-defined, challenging, and relevant to the content of the paper, avoiding artifacts that would make identification trivial, and designing a scalable, automated evaluation metric. On the resulting benchmark, we evaluate five frontier LLMs: Claude Sonnet 4.5, DeepSeek Reasoner v3.1, Gemini 2.5 Pro, GPT 5, and Grok 4. Among these, GPT 5 is the top-performing model, achieving 39.1% identification accuracy when k=10, where k is the number of top-ranked error text candidates generated by the LLM.
Read the full abstract

All authors Sarina Xi, Vishisht Rao, Justin Payan, Nihar B. Shah

Computation and Language

Revisiting Generalization Across Difficulty Levels: It's Not So Easy

Yeganeh Kordi, Nihal V. Nayak, Max Zuo et al. (+2 more)

We investigate how well large language models (LLMs) generalize across different task difficulties, a key question for effective data curation and evaluation. Existing research is mixed regarding whether training on easier or harder data leads to better results, and whether those gains come on easier or harder test data. We address this question by conducting a systematic evaluation of LLMs' generalization across models, datasets, and fine-grained groups of example difficulty. We rank examples in six datasets using the outputs of thousands of different LLMs and Item Response Theory (IRT), a well-established difficulty metric in educational testing. Unlike prior work, our difficulty ratings are therefore determined solely by the abilities of many different LLMs, excluding human opinions of difficulty. With a more objective, larger-scale, and finer-grained analysis, we show that cross-difficulty generalization is often limited; training on either easy or hard data cannot achieve consistent improvements across the full range of difficulties. These results show the importance of having a range of difficulties in both training and evaluation data for LLMs, and that taking shortcuts with respect to difficulty is risky.
Read the full abstract

All authors Yeganeh Kordi, Nihal V. Nayak, Max Zuo, Ilana Nguyen, Stephen H. Bach

Computation and Language

ToolOrchestra: Elevating Intelligence via Efficient Model and Tool Orchestration

Hongjin Su, Shizhe Diao, Ximing Lu et al. (+13 more)

Large language models are powerful generalists, yet solving deep and complex problems such as those of the Humanity's Last Exam (HLE) remains both conceptually challenging and computationally expensive. We show that small orchestrators managing other models and a variety of tools can both push the upper bound of intelligence and improve efficiency in solving difficult agentic tasks. We introduce ToolOrchestra, a method for training small orchestrators that coordinate intelligent tools. ToolOrchestra explicitly uses reinforcement learning with outcome-, efficiency-, and user-preference-aware rewards. Using ToolOrchestra, we produce Orchestrator, an 8B model that achieves higher accuracy at lower cost than previous tool-use agents while aligning with user preferences on which tools are to be used for a given query. On HLE, Orchestrator achieves a score of 37.1%, outperforming GPT-5 (35.1%) while being 2.5x more efficient. On tau2-Bench and FRAMES, Orchestrator surpasses GPT-5 by a wide margin while using only about 30% of the cost. Extensive analysis shows that Orchestrator achieves the best trade-off between performance and cost under multiple metrics, and generalizes robustly to unseen tools. These results demonstrate that composing diverse tools with a lightweight orchestration model is both more efficient and more effective than existing methods, paving the way for practical and scalable tool-augmented reasoning systems.
Read the full abstract

All authors Hongjin Su, Shizhe Diao, Ximing Lu, Mingjie Liu, Jiacheng Xu, Xin Dong, Yonggan Fu, Peter Belcak, Hanrong Ye, Hongxu Yin, Yi Dong, Evelina Bakhturina, Tao Yu, Yejin Choi, Jan Kautz, Pavlo Molchanov

Computation and Language · Cross-listed from cs.CV

G$^2$VLM: Geometry Grounded Vision Language Model with Unified 3D Reconstruction and Spatial Reasoning

Wenbo Hu, Jingli Lin, Yilin Long et al. (+7 more)

Vision-Language Models (VLMs) still lack robustness in spatial intelligence, demonstrating poor performance on spatial understanding and reasoning tasks. We attribute this gap to the absence of a visual geometry learning process capable of reconstructing 3D space from 2D images. We present G$^2$VLM, a geometry grounded vision-language model that bridges two fundamental aspects of spatial intelligence: spatial 3D reconstruction and spatial understanding. G$^2$VLM natively leverages learned 3D visual geometry features to directly predict 3D attributes and enhance spatial reasoning tasks via in-context learning and interleaved reasoning. Our unified design is highly scalable for spatial understanding: it trains on abundant multi-view image and video data, while simultaneously leveraging the benefits of 3D visual priors that are typically only derived from hard-to-collect annotations. Experimental results demonstrate G$^2$VLM is proficient in both tasks, achieving comparable results to state-of-the-art feed-forward 3D reconstruction models and achieving better or competitive results across spatial understanding and reasoning tasks. By unifying a semantically strong VLM with low-level 3D vision tasks, we hope G$^2$VLM can serve as a strong baseline for the community and unlock more future applications, such as 3D scene editing.
Read the full abstract

All authors Wenbo Hu, Jingli Lin, Yilin Long, Yunlong Ran, Lihan Jiang, Yifan Wang, Chenming Zhu, Runsen Xu, Tai Wang, Jiangmiao Pang

Computation and Language

Matrix: Peer-to-Peer Multi-Agent Synthetic Data Generation Framework

Dong Wang, Yang Li, Ansong Ni et al. (+12 more)

Synthetic data has become increasingly important for training large language models, especially when real data is scarce, expensive, or privacy-sensitive. Many such generation tasks require coordinated multi-agent workflows, where specialized agents collaborate to produce data that is higher quality, more diverse, and structurally richer. However, existing frameworks for multi-agent synthesis often depend on a centralized orchestrator, creating scalability bottlenecks, or are hardcoded for specific domains, limiting flexibility. We present Matrix, a decentralized framework that represents both control and data flow as serialized messages passed through distributed queues. This peer-to-peer design eliminates the central orchestrator. Each task progresses independently through lightweight agents, while compute-intensive operations, such as LLM inference or containerized environments, are handled by distributed services. Built on Ray, Matrix scales to tens of thousands of concurrent agentic workflows and provides a modular, configurable design that enables easy adaptation to a wide range of data generation workflows. We evaluate Matrix across diverse synthesis scenarios, such as multi-agent collaborative dialogue, web-based reasoning data extraction, and tool-use trajectory generation in customer service environments. In all cases, Matrix achieves $2$–$15\times$ higher data generation throughput under identical hardware resources, without compromising output quality.
Read the full abstract

All authors Dong Wang, Yang Li, Ansong Ni, Ching-Feng Yeh, Youssef Emad, Xinjie Lei, Liam Robbins, Karthik Padthe, Hu Xu, Xian Li, Asli Celikyilmaz, Ramya Raghavendra, Lifei Huang, Carole-Jean Wu, Shang-Wen Li

Computation and Language

The author is dead, but what if they never lived? A reception experiment on Czech AI- and human-authored poetry

Anna Marklová, Ondřej Vinš, Martina Vokáčová et al. (+1 more)

Large language models are increasingly capable of producing creative texts, yet most studies on AI-generated poetry focus on English – a language that dominates training data. In this paper, we examine the perception of AI- and human-written Czech poetry. We ask if Czech native speakers are able to identify it and how they aesthetically judge it. Participants performed at chance level when guessing authorship (45.8% correct on average), indicating that Czech AI-generated poems were largely indistinguishable from human-written ones. Aesthetic evaluations revealed a strong authorship bias: when participants believed a poem was AI-generated, they rated it as less favorably, even though AI poems were in fact rated equally or more favorably than human ones on average. The logistic regression model uncovered that the more the people liked a poem, the less probable was that they accurately assign the authorship. Familiarity with poetry or literary background had no effect on recognition accuracy. Our findings show that AI can convincingly produce poetry even in a morphologically complex, low-resource (with respect of the training data of AI models) Slavic language such as Czech. The results suggest that readers' beliefs about authorship and the aesthetic evaluation of the poem are interconnected.
Read the full abstract

All authors Anna Marklová, Ondřej Vinš, Martina Vokáčová, Jiří Milička