Peer Revenue Impact factor: volatile

Papers

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

Computer Science Computation and Language

cs.CL Watch Computation and Language

Saturday 25 October

Computation and Language

Benchmarking and Learning Real-World Customer Service Dialogue

Tianhong Gao, Jundong Shen, Jiapeng Wang et al. (+4 more)

Existing benchmarks and training pipelines for industrial intelligent customer service (ICS) remain misaligned with real-world dialogue requirements, overemphasizing verifiable task success while under-measuring subjective service quality and realistic failure modes, leaving a gap between offline gains and deployable dialogue behavior. We close this gap with a benchmark-to-optimization loop: we first introduce OlaBench, an ICS benchmark spanning retrieval-augmented generation, workflow-based systems, and agentic settings, which evaluates service capability, safety, and latency sensitivity; moreover, motivated by OlaBench results showing state-of-the-art LLMs still fall short, we propose OlaMind, which distills reusable reasoning patterns and service strategies from expert dialogues and applies staged exploration–exploitation reinforcement learning with instance-level rubric-aware guidance to improve model capability. OlaMind surpasses GPT-5.2 and Gemini 3 Pro on OlaBench (83.64 vs. 70.58/70.84) and, in online A/B tests, delivers an average +23.67% issue resolution and -6.6% human transfer rate versus the baseline, bridging offline gains to deployment. Together, OlaBench and OlaMind advance ICS systems toward more anthropomorphic, professional, and reliable deployment. The project page and evaluation are available at https://olamind-olabench.github.io.
Read the full abstract

All authors Tianhong Gao, Jundong Shen, Jiapeng Wang, Bei Shi, Ying Ju, Junfeng Yao, Huiyu Yu

Computation and Language · Cross-listed from cs.CV

LOC: A General Language-Guided Framework for Open-Set 3D Occupancy Prediction

Yuhang Gao, Xiang Xiang, Sheng Zhong et al. (+1 more)

Vision-Language Models (VLMs) have shown significant progress in open-set challenges. However, the limited availability of 3D datasets hinders their effective application in 3D scene understanding. We propose LOC, a general language-guided framework adaptable to various occupancy networks, supporting both supervised and self-supervised learning paradigms. For self-supervised tasks, we employ a strategy that fuses multi-frame LiDAR points for dynamic/static scenes, using Poisson reconstruction to fill voids, and assigning semantics to voxels via K-Nearest Neighbor (KNN) to obtain comprehensive voxel representations. To mitigate feature over-homogenization caused by direct high-dimensional feature distillation, we introduce Densely Contrastive Learning (DCL). DCL leverages dense voxel semantic information and predefined textual prompts. This efficiently enhances open-set recognition without dense pixel-level supervision, and our framework can also leverage existing ground truth to further improve performance. Our model predicts dense voxel features embedded in the CLIP feature space, integrating textual and image pixel information, and classifies based on text and semantic similarity. Experiments on the nuScenes dataset demonstrate the method's superior performance, achieving high-precision predictions for known classes and distinguishing unknown classes without additional training data.
Read the full abstract

All authors Yuhang Gao, Xiang Xiang, Sheng Zhong, Guoyou Wang

Computation and Language · Cross-listed from cs.LG

Edit Less, Achieve More: Dynamic Sparse Neuron Masking for Lifelong Knowledge Editing in LLMs

Jinzhe Liu, Junshu Sun, Shufan Shen et al. (+2 more)

Lifelong knowledge editing enables continuous, precise updates to outdated knowledge in large language models (LLMs) without computationally expensive full retraining. However, existing methods often accumulate errors throughout the editing process, causing a gradual decline in both editing accuracy and generalization. To tackle this problem, we propose Neuron-Specific Masked Knowledge Editing (NMKE), a novel fine-grained editing framework that combines neuron-level attribution with dynamic sparse masking. Leveraging neuron functional attribution, we identify two key types of knowledge neurons, with knowledge-general neurons activating consistently across prompts and knowledge-specific neurons activating to specific prompts. NMKE further introduces an entropy-guided dynamic sparse mask, locating relevant neurons to the target knowledge. This strategy enables precise neuron-level knowledge editing with fewer parameter modifications. Experimental results from thousands of sequential edits demonstrate that NMKE outperforms existing methods in maintaining high editing success rates and preserving model general capabilities in lifelong editing.
Read the full abstract

All authors Jinzhe Liu, Junshu Sun, Shufan Shen, Chenxue Yang, Shuhui Wang

Computation and Language

Every Activation Boosted: Scaling General Reasoner to 1 Trillion Open Language Foundation

Ling Team, Ang Li, Ben Liu et al. (+139 more)

We introduce Ling 2.0, a series reasoning-oriented language foundation built upon the principle that every activation boosts reasoning capability. Designed to scale from tens of billions to one trillion parameters under a unified Mixture-of-Experts (MoE) paradigm, Ling 2.0 emphasizes high sparsity, cross-scale consistency, and efficiency guided by empirical scaling laws. The series includes three non-thinking (instruct) models - Ling-mini-2.0, Ling-flash-2.0, and Ling-1T - ranging from 16B to 1T total parameters and achieving up to 7-fold active-compute efficiency compared with dense counterparts. Ling 2.0 integrates coordinated innovations across model architecture, pre-training, post-training, and infrastructure: a high-sparsity MoE with MTP for efficient reasoning, reasoning-oriented data and mid-training CoT activation, reinforcement-based fine-tuning (DFT, Evo-CoT), and full-scale FP8 training with fine-grained heterogeneous pipelines. At the trillion scale, Ling-1T establishes a new Pareto frontier of reasoning accuracy versus computational efficiency, demonstrating that sparse activation, when properly aligned with reasoning objectives, enables scalable and efficient intelligence. Collectively, Ling 2.0 provides a coherent, open, and efficient foundation for advancing future reasoning and thinking models, including the Ring series built upon the same base.
Read the full abstract

All authors Ling Team, Ang Li, Ben Liu, Binbin Hu, Bing Li, Bingwei Zeng, Borui Ye, Caizhi Tang, Changxin Tian, Chao Huang, Chao Zhang, Chen Qian, Chenchen Ju, Chenchen Li, Chengfu Tang, Chilin Fu, Chunshao Ren, Chunwei Wu, Cong Zhang, Cunyin Peng, Dafeng Xu, Daixin Wang, Dalong Zhang, Dingnan Jin, Dingyuan Zhu, Dongke Hu, Fangzheng Zhao, Feifan Wu, Feng Zhu, Gangshan Wang, Haitao Zhang, Hailin Zhao, Hanxiao Zhang, Hanzi Wang, Hao Qian, Haoyi Yu, Heng Zhang, Hongliang Zhang, Hongzhi Luan, Huirong Dong, Huizhong Li, Jia Li, Jia Liu, Jialong Zhu, Jian Sha, Jianping Wei, Jiaolong Yang, Jieyue Ma, Jiewei Wu, Jinjing Huang, Jingyun Tian, Jingyuan Zhang, Jinquan Sun, Juanhui Tu, Jun Liu, Jun Xu, Jun Zhou, Junjie Ou, Junpeng Fang, Kaihong Zhang, Kaiqin Hu, Ke Shi, Kun Tang, Kunlong Chen, Lanyin Mei, Lei Liang, Lei Xu, Libo Zhang, Lin Ju, Lin Yuan, Ling Zhong, Lintao Ma, Lu Liu, Lu Yu, Lun Cai, Meiqi Zhu, Mengying Li, Min Chen, Minghao Xue, Minghong Cai, Mingming Yin, Peijie Jiang, Peilong Zhao, Pingping Liu, Qian Zhao, Qing Cui, Qingxiang Huang, Qingyuan Yang, Quankun Yu, Shaowei Wei, Shijie Lian, Shoujian Zheng, Shun Song, Shungen Zhang, Shuo Zhang, Siyuan Li, Song Liu, Ting Guo, Tong Zhao, Wanli Gu, Weichang Wu, Weiguang Han, Wenjing Fang, Wubin Wang, Xiang Shu, Xiao Shi, Xiaoshun Lan, Xiaolu Zhang, Xiaqing Sun, Xin Zhao, Xingyu Lu, Xiong Xu, Xudong Wang, Xudong Wang, Xuemin Yang, Yajie Yang, Yang Xiang, Yanzhe Li, Yi Zhang, Yilong Wang, Yingxue Li, Yongzhen Guo, Yuzhuo Fu, Yuanyuan Wang, Yue Yang, Yue Yu, Yufeng Deng, Yun Zhang, Yunfei Yu, Yuqi Zhang, Yuxiao He, Zengke Gui, Zhaoxin Huan, Zhaoyang Wang, Zhibo Zhu, Zhihao Wang, Zhiqiang Zhang, Zhoufei Wang, Zihang Zeng, Ziqi Liu, Zitao Xuan, Zuoli Tang

Computation and Language

Gradual Forgetting: Logarithmic Compression for Extending Transformer Context Windows

Billy Dickson and Zoran Tiganj

Most approaches to long-context processing increase the complexity of the transformer's internal architecture by integrating mechanisms such as recurrence or auxiliary memory modules. In this work, we introduce an alternative approach that modifies the input representation itself, rather than the transformer architecture. Inspired by cognitive models of human memory, our method applies a scale-invariant logarithmic compression to the input tokens. The resulting compressed representation is processed by a standard, unmodified transformer, preserving architectural simplicity. We evaluate this approach on the WikiText-103 and PG-19 language modeling benchmarks, showing a reduction in perplexity compared to uncompressed baselines. Moreover, performance improves consistently with longer compressed temporal contexts, showing that input-level logarithmic compression is a simple and effective way to extend a transformer's long-range memory.
Read the full abstract

Computation and Language · Cross-listed from cs.CV

Mitigating Coordinate Prediction Bias from Positional Encoding Failures

Xingjian Tao, Yiwei Wang, Yujun Cai et al. (+3 more)

While Multimodal Large Language Models (MLLMs) excel at general vision-language tasks, precise coordinate prediction remains a significant challenge, particularly as high-resolution inputs cause visual positional encodings (VPEs) to degrade. We demonstrate that these encoding failures do not result in random noise but instead trigger predictable, directional biases, suggesting that models default to internal spatial priors when grounding signals are weak. To counteract this, we introduce Vision-PE Shuffle Guidance (VPSG), a training-free, inference-time correction method. VPSG isolates position-unconditioned tendencies by shuffling VPEs and utilizes this negative evidence to steer digit decoding through a lightweight finite-state machine. Evaluation on the ScreenSpot-Pro benchmark confirms that VPSG effectively rectifies coordinate drift, yielding consistent improvements in localization accuracy across various model scales without any retraining. Our code is available at https://github.com/taoxj2001/VPSG.
Read the full abstract

All authors Xingjian Tao, Yiwei Wang, Yujun Cai, Yihong Luo, Kai Han, Jing Tang

Computation and Language

Generalization or Memorization: Dynamic Decoding for Mode Steering

Xuanming Zhang

Large Language Models (LLMs) exhibit a troubling duality, capable of both remarkable generalization and brittle, verbatim memorization of their training data. This unpredictability undermines their reliability in high-stakes applications. In this work, we propose a unified framework to understand, identify, and control these distinct reasoning modes. First, we introduce a theoretical model based on the Information Bottleneck (IB) principle, formalizing generalization as the learning of a compressed, task-relevant representation and memorization as a failure to compress. Building on this theory, we develop Dynamic Mode Steering (DMS), a novel inference-time algorithm which comprises two components: (1) a lightweight, causally-grounded linear probe that identifies the model's instantaneous reliance on memorization, and (2) a dynamic activation steering mechanism that nudges the model's computation towards pre-identified generalization circuits. We frame DMS as a form of adaptive, self-contrastive decoding. Experiments on reasoning and faithfulness tasks demonstrate that DMS significantly improves logical consistency and factual accuracy, thereby offering a principled approach to enhancing LLM reliability.
Read the full abstract

Computation and Language · Cross-listed from cs.AI

Embracing Trustworthy Brain-Agent Collaboration as Paradigm Extension for Intelligent Assistive Technologies

Yankai Chen, Xinni Zhang, Yifei Zhang et al. (+6 more)

Brain-Computer Interfaces (BCIs) offer a direct communication pathway between the human brain and external devices, holding significant promise for individuals with severe neurological impairments. However, their widespread adoption is hindered by critical limitations, such as low information transfer rates and extensive user-specific calibration. To overcome these challenges, recent research has explored the integration of Large Language Models (LLMs), extending the focus from simple command decoding to understanding complex cognitive states. Despite these advancements, deploying agentic AI faces technical hurdles and ethical concerns. Due to the lack of comprehensive discussion on this emerging direction, this position paper argues that the field is poised for a paradigm extension from BCI to Brain-Agent Collaboration (BAC). We emphasize reframing agents as active and collaborative partners for intelligent assistance rather than passive brain signal data processors, demanding a focus on ethical data handling, model reliability, and a robust human-agent collaboration framework to ensure these systems are safe, trustworthy, and effective.
Read the full abstract

All authors Yankai Chen, Xinni Zhang, Yifei Zhang, Yangning Li, Henry Peng Zou, Chunyu Miao, Weizhi Zhang, Xue Liu, Philip S. Yu

Friday 24 October

Computation and Language · Cross-listed from cs.AI

LLM-Generated Negative News Headlines Dataset: Creation and Benchmarking Against Real Journalism

Olusola Babalola, Bolanle Ojokoh and Olutayo Boyinbode

This research examines the potential of datasets generated by Large Language Models (LLMs) to support Natural Language Processing (NLP) tasks, aiming to overcome challenges related to data acquisition and privacy concerns associated with real-world data. Focusing on negative valence text, a critical component of sentiment analysis, we explore the use of LLM-generated synthetic news headlines as an alternative to real-world data. A specialized corpus of negative news headlines was created using tailored prompts to capture diverse negative sentiments across various societal domains. The synthetic headlines were validated by expert review and further analyzed in embedding space to assess their alignment with real-world negative news in terms of content, tone, length, and style. Key metrics such as correlation with real headlines, perplexity, coherence, and realism were evaluated. The synthetic dataset was benchmarked against two sets of real news headlines using evaluations including the Comparative Perplexity Test, Comparative Readability Test, Comparative POS Profiling, BERTScore, and Comparative Semantic Similarity. Results show the generated headlines match real headlines with the only marked divergence being in the proper noun score of the POS profile test.
Read the full abstract

Computation and Language · Cross-listed from cs.AI

Multimodal Detection of Fake Reviews using BERT and ResNet-50

Suhasnadh Reddy Veluru, Sai Teja Erukude and Viswa Chaitanya Marella

In the current digital commerce landscape, user-generated reviews play a critical role in shaping consumer behavior, product reputation, and platform credibility. However, the proliferation of fake or misleading reviews often generated by bots, paid agents, or AI models poses a significant threat to trust and transparency within review ecosystems. Existing detection models primarily rely on unimodal, typically textual, data and therefore fail to capture semantic inconsistencies across different modalities. To address this gap, a robust multimodal fake review detection framework is proposed, integrating textual features encoded with BERT and visual features extracted using ResNet-50. These representations are fused through a classification head to jointly predict review authenticity. To support this approach, a curated dataset comprising 21,142 user-uploaded images across food delivery, hospitality, and e-commerce domains was utilized. Experimental results indicate that the multimodal model outperforms unimodal baselines, achieving an F1-score of 0.934 on the test set. Additionally, the confusion matrix and qualitative analysis highlight the model's ability to detect subtle inconsistencies, such as exaggerated textual praise paired with unrelated or low-quality images, commonly found in deceptive content. This study demonstrates the critical role of multimodal learning in safeguarding digital trust and offers a scalable solution for content moderation across various online platforms.
Read the full abstract

Computation and Language

Confidence is Not Competence

Debdeep Sanyal, Manya Pandey, Dhruv Kumar et al. (+2 more)

Large language models (LLMs) often exhibit a puzzling disconnect between their asserted confidence and actual problem-solving competence. We offer a mechanistic account of this decoupling by analyzing the geometry of internal states across two phases - pre-generative assessment and solution execution. A simple linear probe decodes the internal "solvability belief" of a model, revealing a well-ordered belief axis that generalizes across model families and across math, code, planning, and logic tasks. Yet, the geometries diverge - although belief is linearly decodable, the assessment manifold has high linear effective dimensionality as measured from the principal components, while the subsequent reasoning trace evolves on a much lower-dimensional manifold. This sharp reduction in geometric complexity from thought to action mechanistically explains the confidence-competence gap. Causal interventions that steer representations along the belief axis leave final solutions unchanged, indicating that linear nudges in the complex assessment space do not control the constrained dynamics of execution. We thus uncover a two-system architecture - a geometrically complex assessor feeding a geometrically simple executor. These results challenge the assumption that decodable beliefs are actionable levers, instead arguing for interventions that target the procedural dynamics of execution rather than the high-level geometry of assessment.
Read the full abstract

All authors Debdeep Sanyal, Manya Pandey, Dhruv Kumar, Saurabh Deshpande, Murari Mandal

Computation and Language · Cross-listed from cs.SE

VisCoder2: Building Multi-Language Visualization Coding Agents

Yuansheng Ni, Songcheng Cai, Xiangchao Chen et al. (+8 more)

Large language models (LLMs) have recently enabled coding agents capable of generating, executing, and revising visualization code. However, existing models often fail in practical workflows due to limited language coverage, unreliable execution, and lack of iterative correction mechanisms. Progress has been constrained by narrow datasets and benchmarks that emphasize single-round generation and single-language tasks. To address these challenges, we introduce three complementary resources for advancing visualization coding agents. VisCode-Multi-679K is a large-scale, supervised dataset containing 679K validated and executable visualization samples with multi-turn correction dialogues across 12 programming languages. VisPlotBench is a benchmark for systematic evaluation, featuring executable tasks, rendered outputs, and protocols for both initial generation and multi-round self-debug. Finally, we present VisCoder2, a family of multi-language visualization models trained on VisCode-Multi-679K. Experiments show that VisCoder2 significantly outperforms strong open-source baselines and approaches the performance of proprietary models like GPT-4.1, with further gains from iterative self-debug, reaching 82.4% overall execution pass rate at the 32B scale, particularly in symbolic or compiler-dependent languages.
Read the full abstract

All authors Yuansheng Ni, Songcheng Cai, Xiangchao Chen, Jiarong Liang, Zhiheng Lyu, Jiaqi Deng, Kai Zou, Ping Nie, Fei Yuan, Xiang Yue, Wenhu Chen

Computation and Language · Cross-listed from cs.LG

Combining Textual and Structural Information for Premise Selection in Lean

Job Petrovčič, David Eliecer Narvaez Denis and Ljupčo Todorovski

Premise selection is a key bottleneck for scaling theorem proving in large formal libraries. Yet existing language-based methods often treat premises in isolation, ignoring the web of dependencies that connects them. We present a graph-augmented approach that combines dense text embeddings of Lean formalizations with graph neural networks over a heterogeneous dependency graph capturing both state-premise and premise-premise relations. On the LeanDojo Benchmark, our method outperforms the ReProver language-based baseline by over 25% across standard retrieval metrics. These results suggest that relational information is beneficial for premise selection.
Read the full abstract

Computation and Language · Cross-listed from cs.LG

LLM4Delay: Flight Delay Prediction via Cross-Modality Adaptation of Large Language Models and Aircraft Trajectory Representation

Thaweerath Phisannupawong, Joshua Julian Damanik and Han-Lim Choi

Flight delay prediction has become a key focus in air traffic management (ATM), as delays reflect inefficiencies in the system. This paper proposes LLM4Delay, a large language model (LLM)-based framework for predicting flight delays from the perspective of air traffic controllers monitoring aircraft after they enter the terminal maneuvering area (TMA). LLM4Delay is designed to integrate textual aeronautical information, including flight data, weather reports, and aerodrome notices, together with multiple trajectories that model airspace conditions, forming a comprehensive delay-relevant context. By jointly leveraging comprehensive textual and trajectory contexts via instance-level projection, an effective cross-modality adaptation strategy that maps multiple instance-level trajectory representations into the language modality, the framework improves delay prediction accuracy. LLM4Delay demonstrates superior performance compared to existing ATM frameworks and prior time-series-to-language adaptation methods. This highlights the complementary roles of textual and trajectory data while leveraging knowledge from both the pretrained trajectory encoder and the pretrained LLM. The proposed framework enables continuous updates to predictions as new information becomes available, indicating potential operational relevance.
Read the full abstract

Computation and Language · Cross-listed from cs.LG

NUM2EVENT: Interpretable Event Reasoning from Numerical time-series

Ninghui Feng and Yiyan Qi

Large language models (LLMs) have recently demonstrated impressive multimodal reasoning capabilities, yet their understanding of purely numerical time-series signals remains limited. Existing approaches mainly focus on forecasting or trend description, without uncovering the latent events that drive numerical changes or explaining the reasoning process behind them. In this work, we introduce the task of number-to-event reasoning and decoding, which aims to infer interpretable structured events from numerical inputs, even when current text is unavailable. To address the data scarcity and semantic alignment challenges, we propose a reasoning-aware framework that integrates an agent-guided event extractor (AGE), a marked multivariate Hawkes-based synthetic generator (EveDTS), and a two-stage fine-tuning pipeline combining a time-series encoder with a structured decoder. Our model explicitly reasons over numerical changes, generates intermediate explanations, and outputs structured event hypotheses. Experiments on multi-domain datasets show that our method substantially outperforms strong LLM baselines in event-level precision and recall. These results suggest a new direction for bridging quantitative reasoning and semantic understanding, enabling LLMs to explain and predict events directly from numerical dynamics.
Read the full abstract

Computation and Language · Cross-listed from cs.CR

Jailbreak Mimicry: Automated Discovery of Narrative-Based Jailbreaks for Large Language Models

Pavlos Ntais

Large language models (LLMs) remain vulnerable to sophisticated prompt engineering attacks that exploit contextual framing to bypass safety mechanisms, posing significant risks in cybersecurity applications. We introduce Jailbreak Mimicry, a systematic methodology for training compact attacker models to automatically generate narrative-based jailbreak prompts in a one-shot manner. Our approach transforms adversarial prompt discovery from manual craftsmanship into a reproducible scientific process, enabling proactive vulnerability assessment in AI-driven security systems. Developed for the OpenAI GPT-OSS-20B Red-Teaming Challenge, we use parameter-efficient fine-tuning (LoRA) on Mistral-7B with a curated dataset derived from AdvBench, achieving an 81.0% Attack Success Rate (ASR) against GPT-OSS-20B on a held-out test set of 200 items. Cross-model evaluation reveals significant variation in vulnerability patterns: our attacks achieve 66.5% ASR against GPT-4, 79.5% on Llama-3 and 33.0% against Gemini 2.5 Flash, demonstrating both broad applicability and model-specific defensive strengths in cybersecurity contexts. This represents a 54x improvement over direct prompting (1.5% ASR) and demonstrates systematic vulnerabilities in current safety alignment approaches. Our analysis reveals that technical domains (Cybersecurity: 93% ASR) and deception-based attacks (Fraud: 87.8% ASR) are particularly vulnerable, highlighting threats to AI-integrated threat detection, malware analysis, and secure systems, while physical harm categories show greater resistance (55.6% ASR). We employ automated harmfulness evaluation using Claude Sonnet 4, cross-validated with human expert assessment, ensuring reliable and scalable evaluation for cybersecurity red-teaming. Finally, we analyze failure mechanisms and discuss defensive strategies to mitigate these vulnerabilities in AI for cybersecurity.
Read the full abstract

Computation and Language

Compositional Bias Control in Large Language Models: Preference Learning Fails, Supervision Succeeds

Atij Mahesh

Large Language Models (LLMs) still produce gender-stereotyped language even in occupation-neutral contexts that reflect deep societal biases (Rudinger et al., 2018). To address this, prior work has proposed prompting, constrained decoding (Dathathri et al., 2020; Zhou et al., 2024), post-processing, and fine-tuning-based alignment (Rafailov et al., 2023; Ravfogel et al., 2022). However, the comparative efficacy and learning dynamics remain little understood. We report a comparative analysis of six control techniques for bias mitigation: prompt-only, generate-and-filter, DFA-based Ctrl-G decoding, Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), and Iterative Nullspace Projection (INLP). We evaluate each method on a compositional constraint task. This task requires generating sentences that contain at least one agentic and one communal descriptor for each of the twenty Winogender-derived occupations. We quantify trade-offs between control strength and naturalness with evaluations of constraint compliance, lexical diversity, and fluency. Our results reveal key contrasts among the methods: SFT achieves 99.87 +- 0.15% compliance and high lexical diversity, while DPO, despite similar training stability, fails at 4.53 +- 0.82%. Ctrl-G guarantees perfect compliance, but at the cost of severely reduced fluency and diversity. Preference-based learning fundamentally differs: it cannot satisfy compositional constraints, as binary preference signals encode ranking, not logical conjunctions. Only explicit positive supervision enables mitigation of compositional biases; preference-based alignment fails to generalize logical structures, underscoring the limitations of preference learning and the necessity of explicit supervision for fair and fluent controlled generation.
Read the full abstract

Computation and Language · Cross-listed from cs.LG

Agentic Reinforcement Learning for Real-World Code Repair

Siyu Zhu, Anastasiya Karpovich, Albert Chen et al. (+6 more)

We tackle the challenge of training reliable code-fixing agents in real repositories, where complex builds and shifting dependencies make evaluation unstable. We developed a verifiable pipeline with success defined as post-fix build validation and improved reproducibility across ~1K real issues by pinning dependencies and disabling automatic upgrades. Building on this, we introduced a scalable simplified pipeline for large-scale reinforcement learning (RL). Using this setup, we supervised fine-tuned Qwen3-32B in the full pipeline and applied RL on top of the SFT model in the simplified environment. The SFT model distilled from GPT-4.1 trajectories performs on par while being 56x smaller, and RL added 7-20% absolute gains under matched train-test conditions. "Thinking mode" was on par or worse in our experiments. Both SFT and RL models failed to generalize across environments, highlighting the importance of matching train-test environments for building reliable real-world code-fixing agents.
Read the full abstract

All authors Siyu Zhu, Anastasiya Karpovich, Albert Chen, Jessica Koscheka, Shailesh Jannu, Di Wen, Yuqing Zhu, Rohit Jain, Alborz Geramifard

Computation and Language · Cross-listed from cs.IR

A Benchmark for Open-Domain Numerical Fact-Checking Enhanced by Claim Decomposition

V Venktesh, Deepali Prabhu and Avishek Anand

Fact-checking numerical claims is critical as the presence of numbers provide mirage of veracity despite being fake potentially causing catastrophic impacts on society. The prior works in automatic fact verification do not primarily focus on natural numerical claims. A typical human fact-checker first retrieves relevant evidence addressing the different numerical aspects of the claim and then reasons about them to predict the veracity of the claim. Hence, the search process of a human fact-checker is a crucial skill that forms the foundation of the verification process. Emulating a real-world setting is essential to aid in the development of automated methods that encompass such skills. However, existing benchmarks employ heuristic claim decomposition approaches augmented with weakly supervised web search to collect evidences for verifying claims. This sometimes results in less relevant evidences and noisy sources with temporal leakage rendering a less realistic retrieval setting for claim verification. Hence, we introduce QuanTemp++: a dataset consisting of natural numerical claims, an open domain corpus, with the corresponding relevant evidence for each claim. The evidences are collected through a claim decomposition process approximately emulating the approach of human fact-checker and veracity labels ensuring there is no temporal leakage. Given this dataset, we also characterize the retrieval performance of key claim decomposition paradigms. Finally, we observe their effect on the outcome of the verification pipeline and draw insights. The code for data pipeline along with link to data can be found at https://github.com/VenkteshV/QuanTemp_Plus
Read the full abstract

Computation and Language

Emotions Where Art Thou: Understanding and Characterizing the Emotional Latent Space of Large Language Models

Benjamin Reichman, Adar Avsian and Larry Heck

This work investigates how large language models (LLMs) internally represent emotion by analyzing the geometry of their hidden-state space. The paper identifies a low-dimensional emotional manifold and shows that emotional representations are directionally encoded, distributed across layers, and aligned with interpretable dimensions. These structures are stable across depth and generalize to eight real-world emotion datasets spanning five languages. Cross-domain alignment yields low error and strong linear probe performance, indicating a universal emotional subspace. Within this space, internal emotion perception can be steered while preserving semantics using a learned intervention module, with especially strong control for basic emotions across languages. These findings reveal a consistent and manipulable affective geometry in LLMs and offer insight into how they internalize and process emotion.
Read the full abstract

Computation and Language

ATLAS: Adaptive Transfer Scaling Laws for Multilingual Pretraining, Finetuning, and Decoding the Curse of Multilinguality

Shayne Longpre, Sneha Kudugunta, Niklas Muennighoff et al. (+6 more)

Scaling laws research has focused overwhelmingly on English – yet the most prominent AI models explicitly serve billions of international users. In this work, we undertake the largest multilingual scaling laws study to date, totaling 774 multilingual training experiments, spanning 10M-8B model parameters, 400+ training languages and 48 evaluation languages. We introduce the Adaptive Transfer Scaling Law (ATLAS) for both monolingual and multilingual pretraining, which outperforms existing scaling laws' out-of-sample generalization often by more than 0.3 R^2. Our analyses of the experiments shed light on multilingual learning dynamics, transfer properties between languages, and the curse of multilinguality. First, we derive a cross-lingual transfer matrix, empirically measuring mutual benefit scores between 38 x 38=1444 language pairs. Second, we derive a language-agnostic scaling law that reveals how to optimally scale model size and data when adding languages without sacrificing performance. Third, we identify the computational crossover points for when to pretrain from scratch versus finetune from multilingual checkpoints. We hope these findings provide the scientific foundation for democratizing scaling laws across languages, and enable practitioners to efficiently scale models – beyond English-first AI.
Read the full abstract

All authors Shayne Longpre, Sneha Kudugunta, Niklas Muennighoff, I-Hung Hsu, Isaac Caswell, Alex Pentland, Sercan Arik, Chen-Yu Lee, Sayna Ebrahimi

Computation and Language

Penalizing Length: Uncovering Systematic Bias in Quality Estimation Metrics

Yilin Zhang, Wenda Xu, Zhongtao Liu et al. (+2 more)

Quality Estimation (QE) metrics are vital in machine translation for reference-free evaluation and increasingly serve as selection criteria in data filtering and candidate reranking. However, the prevalence and impact of length bias in QE metrics have been underexplored. Through a systematic study of top-performing learned and LLM-as-a-Judge QE metrics across 10 diverse language pairs, we reveal two critical length effects: First, QE metrics consistently over-predict errors with increasing translation length, even for high-quality, error-free texts. Second, when candidates of comparable quality are available for the same source text, learned QE metrics tend to favor shorter translations, whereas LLM-as-a-Judge metrics range from being approximately length-neutral to favoring longer translations. These metric-dependent length effects can favor or penalize translations based on length rather than quality and can propagate into downstream pipelines that rely on QE signals for data selection or system optimization. We trace the root cause of MetricX-24 QE's cumulative-length penalty to skewed supervision distributions, in which longer error-free examples are underrepresented in training data. As a diagnostic intervention, we apply length normalization during training and show that this simple modification effectively decouples error prediction from sequence length, yielding more reliable QE signals across translations of varying length.
Read the full abstract

All authors Yilin Zhang, Wenda Xu, Zhongtao Liu, Tetsuji Nakagawa, Markus Freitag

Computation and Language

Toward Understanding the Transferability of Adversarial Suffixes in Large Language Models

Sarah Ball, Niki Hasrati, Alexander Robey et al. (+4 more)

Discrete optimization-based jailbreaking attacks on large language models aim to generate short, nonsensical suffixes that, when appended onto input prompts, elicit disallowed content. Notably, these suffixes are often transferable – succeeding on prompts and models for which they were never optimized. And yet, despite the fact that transferability is surprising and empirically well-established, the field lacks a rigorous analysis of when and why transfer occurs. To fill this gap, we identify three statistical properties that strongly correlate with transfer success across numerous experimental settings: (1) how much a prompt without a suffix activates a model's internal refusal direction, (2) how strongly a suffix induces a push away from this direction, and (3) how large these shifts are in directions orthogonal to refusal. On the other hand, we find that prompt semantic similarity only weakly correlates with transfer success. These findings lead to a more fine-grained understanding of transferability, which we use in interventional experiments to showcase how our statistical analysis can translate into practical improvements in attack success.
Read the full abstract

All authors Sarah Ball, Niki Hasrati, Alexander Robey, Avi Schwarzschild, Frauke Kreuter, Zico Kolter, Andrej Risteski

Computation and Language · Cross-listed from math.ST

Optimal Detection for Language Watermarks with Pseudorandom Collision

T. Tony Cai, Xiang Li, Qi Long et al. (+2 more)

Text watermarking plays a crucial role in ensuring the traceability and accountability of large language model (LLM) outputs and mitigating misuse. While promising, most existing methods assume perfect pseudorandomness. In practice, repetition in generated text induces collisions that create structured dependence, compromising Type I error control and invalidating standard analyses. We introduce a statistical framework that captures this structure through a hierarchical two-layer partition. At its core is the concept of minimal units – the smallest groups treatable as independent across units while permitting dependence within. Using minimal units, we define a non-asymptotic efficiency measure and cast watermark detection as a minimax hypothesis testing problem. Applied to Gumbel-max and inverse-transform watermarks, our framework produces closed-form optimal rules. It explains why discarding repeated statistics often improves performance and shows that within-unit dependence must be addressed unless degenerate. Both theory and experiments confirm improved detection power with rigorous Type I error control. These results provide the first principled foundation for watermark detection under imperfect pseudorandomness, offering both theoretical insight and practical guidance for reliable tracing of model outputs.
Read the full abstract

All authors T. Tony Cai, Xiang Li, Qi Long, Weijie J. Su, Garrett G. Wen

Computation and Language · Cross-listed from cs.SI

AI-Mediated Communication Reshapes Social Structure in Opinion-Diverse Groups

Faria Huq, Elijah L. Claggett and Hirokazu Shirado

Group segregation or cohesion can emerge from micro-level communication, and AI-assisted messaging may shape this process. Here, we report a preregistered online experiment (N = 557 across 60 sessions) in which participants discussed controversial political topics over multiple rounds and could freely change groups. Some participants received real-time message suggestions from a large language model (LLM), either personalized to their stance (individual assistance) or incorporating their group members' perspectives (relational assistance). We find that small variations in AI-mediated communication cascade into macro-level differences in group composition. Participants with individual assistance send more messages and show greater stance-based clustering, whereas those with relational assistance use more receptive language and form more heterogeneous ties. Hybrid expressive processes-jointly produced by humans and AI-can reshape collective organization. The patterns of structural division and cohesion depend on how AI incorporates users' interaction context.
Read the full abstract

Computation and Language

Uncovering the Persuasive Fingerprint of LLMs in Jailbreaking Attacks

Havva Alizadeh Noughabi, Julien Serbanescu, Fattane Zarrinkalam et al. (+1 more)

Despite recent advances, Large Language Models remain vulnerable to jailbreak attacks that bypass alignment safeguards and elicit harmful outputs. While prior research has proposed various attack strategies differing in human readability and transferability, little attention has been paid to the linguistic and psychological mechanisms that may influence a model's susceptibility to such attacks. In this paper, we examine an interdisciplinary line of research that leverages foundational theories of persuasion from the social sciences to craft adversarial prompts capable of circumventing alignment constraints in LLMs. Drawing on well-established persuasive strategies, we hypothesize that LLMs, having been trained on large-scale human-generated text, may respond more compliantly to prompts with persuasive structures. Furthermore, we investigate whether LLMs themselves exhibit distinct persuasive fingerprints that emerge in their jailbreak responses. Empirical evaluations across multiple aligned LLMs reveal that persuasion-aware prompts significantly bypass safeguards, demonstrating their potential to induce jailbreak behaviors. This work underscores the importance of cross-disciplinary insight in addressing the evolving challenges of LLM safety. The code and data are available.
Read the full abstract

All authors Havva Alizadeh Noughabi, Julien Serbanescu, Fattane Zarrinkalam, Ali Dehghantanha

Computation and Language · Cross-listed from cs.AI

Performance Trade-offs of Optimizing Small Language Models for E-Commerce

Josip Tomo Licardo and Nikola Tankovic

Large Language Models (LLMs) offer state-of-the-art performance in natural language understanding and generation tasks. However, the deployment of leading commercial models for specialized tasks, such as e-commerce, is often hindered by high computational costs, latency, and operational expenses. This paper investigates the viability of smaller, open-weight models as a resource-efficient alternative. We present a methodology for optimizing a one-billion-parameter Llama 3.2 model for multilingual e-commerce intent recognition. The model was fine-tuned using Quantized Low-Rank Adaptation (QLoRA) on a synthetically generated dataset designed to mimic real-world user queries. Subsequently, we applied post-training quantization techniques, creating GPU-optimized (GPTQ) and CPU-optimized (GGUF) versions. Our results demonstrate that the specialized 1B model achieves 99% accuracy, matching the performance of the significantly larger GPT-4.1 model. A detailed performance analysis revealed critical, hardware-dependent trade-offs: while 4-bit GPTQ reduced VRAM usage by 41%, it paradoxically slowed inference by 82% on an older GPU architecture (NVIDIA T4) due to dequantization overhead. Conversely, GGUF formats on a CPU achieved a speedup of up to 18x in inference throughput and a reduction of over 90% in RAM consumption compared to the FP16 baseline. We conclude that small, properly optimized open-weight models are not just a viable but a more suitable alternative for domain-specific applications, offering state-of-the-art accuracy at a fraction of the computational cost.
Read the full abstract

Computation and Language · Cross-listed from cs.LG

Parallel Sampling from Masked Diffusion Models via Conditional Independence Testing

Iskander Azangulov, Teodora Pandeva, Niranjani Prasad et al. (+2 more)

Masked diffusion models (MDMs) offer a compelling alternative to autoregressive models (ARMs) for discrete text generation because they enable parallel token sampling, rather than sequential, left-to-right generation. This means potentially much faster inference. However, effective parallel sampling faces two competing requirements: (i) simultaneously updated tokens must be conditionally independent, and (ii) updates should prioritise high-confidence predictions. These goals conflict because high-confidence predictions often cluster and depend on each other, opportunities for parallel updates. We present PUNT, a model-agnostic sampler that reconciles this trade-off. Our method identifies token dependencies and removes lower-confidence tokens from conflicting groups. This produces sets of indices for unmasking that satisfy both independence and confidence criteria. Our approach ensures improved parallel unmasking through approximate conditional independence testing. Our experiments show that PUNT delivers a superior trade-off between accuracy and compute when compared to other strong training-free baselines, especially for generation of longer sequences. On the IFEval benchmark, it achieves up to 16% higher accuracy over baseline methods, including sequential generation (one-by-one). These gains hold across different values of hyperparameters, mitigating the need for brittle hyperparameter tuning. Moreover, we observe that PUNT induces an emergent hierarchical generation strategy, where the model first establishes high-level paragraph structure before local refinement, suggesting a planning-like generation process that contributes to strong alignment performance.
Read the full abstract

All authors Iskander Azangulov, Teodora Pandeva, Niranjani Prasad, Javier Zazo, Sushrut Karmalkar

Computation and Language

A Stylometric Application of Large Language Models

Harrison F. Stropkay, Jiayi Chen, Mohammad J. Latifi et al. (+2 more)

We show that large language models (LLMs) can be used to distinguish the writings of different authors. Specifically, an individual GPT-2 model, trained from scratch on the works of one author, will predict held-out text from that author more accurately than held-out text from other authors. We suggest that, in this way, a model trained on one author's works embodies the unique writing style of that author. We first demonstrate our approach on books written by eight different (known) authors. We also use this approach to confirm R. P. Thompson's authorship of the well-studied 15th book of the Oz series, originally attributed to F. L. Baum.
Read the full abstract

All authors Harrison F. Stropkay, Jiayi Chen, Mohammad J. Latifi, Daniel N. Rockmore, Jeremy R. Manning

Computation and Language · Cross-listed from cs.LG

Transformer Based Linear Attention with Optimized GPU Kernel Implementation

Armin Gerami and Ramani Duraiswami

The original softmax-based attention mechanism (regular attention) in the extremely successful Transformer architecture computes attention between $N$ tokens, each embedded in a $D$-dimensional head, with a time complexity of $O(N^2D)$. Given the success of Transformers, improving their runtime during both training and inference is a popular research area. One such approach is the introduction of the linear attention (LA) mechanisms, which offers a linear time complexity of $O(ND^2)$ and have demonstrated comparable accuracy to regular attention. However, LA in practice lags behind its theoretical efficiency. We propose a novel method for LA's forward and backward passes, along with a highly-optimized CUDA implementation. Our approach outperforms the state-of-the-art by 3.3 times in speed and reduces memory consumption by 3.6 times. We validate these improvements in both single-layer and end-to-end settings by training a 1.4 billion parameter language model, which demonstrates similar expressivity to regular attention on major reasoning benchmarks.
Read the full abstract