Computation and Language
LongAttnComp: Cross-Family Context Compression for Long-Context Reasoning
Mengmeng Ji, Ravi Shanker Raju, Jonathan Lingjie Li et al. (+1 more)
As real-world applications increasingly require processing inputs of 100k+
tokens, the gap between context length and inference efficiency has become a critical bottleneck. Context compression offers a way to reduce prefill costs while preserving task accuracy. However, existing training-free
attention-based methods leave substantial gaps in demanding long-context tasks such as code
reasoning. We present LongAttnComp, a long-context adaptation of AttnComp that fine-tunes a lightweight cross-
attention scoring layer and introduces tokenlevel chunking, a
token-budget top-p
algorithm, positional reordering, and a formatagnostic query parser. We further design a two-stage fine-tuning recipe for the compressor: Stage 1 builds a general
retrieval foundation from NIAH-style data, and Stage 2 extends it with multi-hop and
reasoning data for broader long-context task coverage. On InfiniteBench Code-Debug, LongAttnComp matches or exceeds full-context accuracy, substantially outperforms training-free baselines, and transfers across four target models from three families. On LongBench v2, the two-stage recipe largely closes the Stage 1 gap on multi-document
reasoning while preserving Code-Debug performance.
Read the full abstract ▾
All authors Mengmeng Ji, Ravi Shanker Raju, Jonathan Lingjie Li, Chen Wu
Computation and Language
DiffuSent: Towards a Unified Diffusion Framework for Aspect-Based Sentiment Analysis
Shu Long, Yanglei Gan and Xuchuan Zhou
Aspect-Based Sentiment Analysis (ABSA) encompasses seven distinct subtasks, each focusing on different extracted elements. Despite the proven success of generative models in unified aspect sentiment analysis, existing approaches often rely on auto-regressive
token-by-
token generation without grasping the whole information of the aspect and opinion terms, resulting in boundary insensitivity, particularly in context of multi-word aspect and opinion terms. To address these issues, we present DiffuSent, a non-auto-regressive
diffusion framework that systematically formulates all ABSA subtasks as boundary denoising
diffusion processes, progressively refining boundaries over noisy states. Furthermore, we introduce a contrastive denoising training strategy which effectively address duplicate predictions with subtle variations introduced by
diffusion process. Extensive experiments across 28 settings (7 subtasks x 4
datasets) demonstrate that DiffuSent achieves delivers consistent improvements over the strongest generative and span-based systems. DiffuSent exhibits notable gains on multi-word triplets, achieving an average improvement of +2.48 F1, and maintains robust extraction accuracy in sentences containing multiple sentiment triplets. Moreover, the non-auto-regressive decoding enables substantial efficiency benefits, reaching up to 181 times faster inference than auto-regressive generative baselines
Read the full abstract ▾
Computation and Language
TukaBench: A Culturally Grounded Jailbreak Benchmark for African Languages
Victor Akinode, Senyu Li, Wassim Hamidouche et al. (+3 more)
Safety evaluation of Large
Language Models (LLMs) remains heavily English-centric, leaving Low-Resource
Languages (LRLs), particularly African ones, critically underexplored. We introduce TUKABENCH, a jailbreak
benchmark for seven African
languages that extends JailbreakBench (JBB) beyond direct translation through four settings: human translation of JBB
prompts, English adaptation to African contexts followed by human translation, human-curated
prompts validated through interactions with GPT-5.2, and code-switched
prompts combining English and African
languages, isolating the effect of
language, cultural grounding, and
prompt evasiveness on model safety. Across closed and open models, prompting in African
languages reduces refusal relative to English, with culturally adapted
prompts leading to least refusal. The evaluation also surfaces two structural limitations: model comprehension failures and reduced LLM-as-a-judge reliability in LRLs. To capture the first, we introduce Deflection alongside Refused and Jailbroken; to assess the second, we validate outputs with human annotations, showing that judge-human agreement drops in lower-resource
languages and less commonly supported scripts.
Read the full abstract ▾
All authors Victor Akinode, Senyu Li, Wassim Hamidouche, Waqas Zamir, Inbal Becker-Reshef, David Ifeoluwa Adelani
Computation and Language
SkillAdaptor: Self-Adapting Skills for LLM Agents from Trajectories
Zhuoyun Yu, Xin Xie, Wuguannan Yao et al. (+4 more)
Large
language model (LLM)
agents increasingly rely on reusable external skills to solve long-horizon interactive tasks. Existing training-free skill adaptation pipelines usually update skills from full trajectories or session-level feedback, which makes failure attribution coarse and often produces unstable or overly broad revisions. We propose SkillAdaptor, a training-free step-level skill adaptation framework with explicit failure attribution, and it can plug into OpenClaw-class
agent harnesses. Given a failed trajectory, SkillAdaptor identifies a first actionable fault step, links responsibility to candidate skills, and applies targeted updates under explicit acceptance checks while keeping the backbone frozen. We evaluate on WebShop, PinchBench, and Claw-Eval with Kimi-K2.5, GLM-5, and GPT-5.2. SkillAdaptor improves over no-skill and skill-adaptation baselines on all three suites, with the largest single-metric improvements of +1.5 points on PinchBench Avg Score%, +1.8 on Claw-Eval Avg Score, and +1.7 on WebShop success rate. These results indicate that step-level attribution supports more stable and auditable training-free skill maintenance\footnote{The code will be released at https://github.com/zjunlp/SkillAdaptor.}.
Read the full abstract ▾
All authors Zhuoyun Yu, Xin Xie, Wuguannan Yao, Chenxi Wang, Lei Liang, Xiang Qi, Shumin Deng
Computation and Language
Med-HEAL: Analyzing and Mitigating Hallucinations in Medical LLMs with Hallucination-Aware In-Context Learning
Yiming Liao, Zeno Franco, Jose Eduardo Lizarraga Mazaba et al. (+1 more)
Hallucinations in medical large
language models (LLMs) pose serious risks for clinical decision support, particularly when models must reason over complex electronic health records (EHRs). However, existing
benchmarks often lack a realistic clinical context and provide limited insight into how
hallucinations can be mitigated in practice. We introduce Med-HEAL, a framework for systematically identifying, analyzing, and mitigating
hallucinations in medical LLMs using clinically grounded data. Building on the EHRNoteQA
benchmark derived from MIMIC-IV discharge summaries, we construct a
hallucination dataset by evaluating BioMistral-7B on open-ended clinical question answering tasks. Model outputs are labeled through a dual evaluation pipeline that combines LLM-as-a-Judge assessment (GPT-4o) with human
auditing by medical student reviewers, producing correctness judgments and annotations of
reasoning errors via a custom web-based evaluation system. We then leverage this
dataset to investigate mitigation strategies: a self-critique pipeline, in which the test model reviews its own answers to detect potential errors and regenerates responses for flagged cases, and
retrieval-augmented in-context learning (RA-ICL), which exposes the model to hallucinated and corrected examples. Experiments across five open-source LLMs-BioMistral, Llama-3.1, DeepSeek, Qwen2.5, and Qwen3, show that the self-critique strategy improves accuracy for three of five models (p < 0.05) without requiring parameter updates. Med-HEAL provides both a reusable
hallucination dataset and a practical framework for studying and mitigating
hallucinations in medical LLMs, supporting safer deployment of AI systems in clinical environments. Our code and data are publicly available at https://github.com/yimingliao-blad/med-heal.git.
Read the full abstract ▾
All authors Yiming Liao, Zeno Franco, Jose Eduardo Lizarraga Mazaba, Keke Chen
Computation and Language
Challenger at MultiPRIDE: Is It Hate Speech or Reclaimed?
Hadi Bayrami Asl Tekanlou, Mahdi Bakhtiyarzadeh and Jafar Razmara
The spread of hate
speech has become increasingly harmful in modern digital environments, particularly on social networking platforms. While recent advances have shown promising results in automatic hate
speech detection, a key challenge remains: distinguishing genuine hate
speech from reclaimed
language. Accurate labeling is difficult due to the nuanced and context-dependent nature of reclaimed expressions. In this paper, we present a simple and interpretable approach for distinguishing hate
speech from reclaimed
language, developed for the MultiPride Shared Task. Our method generates dense semantic text
embeddings and incorporates a label-noise filtering stage using Cleanlab with logistic regression, followed by a Multi-layer Perceptron (MLP) neural network for final classification. The system is designed to operate under limited computational resources while maintaining strong performance. We evaluate our approach using precision, recall, and F1-score, including macro-averaged metrics. Experimental results demonstrate robust performance despite extreme class imbalance in the
dataset. Overall, the findings highlight the potential for further improvements through larger
embedding models and more advanced preprocessing techniques while preserving interpretability.
Read the full abstract ▾
Computation and Language
Don't Read Everything: A Curvature-Conditioned Query for Linear Attention
Dong Le, Thong Nguyen, Cong-Duy Nguyen et al. (+1 more)
Linear
attention reduces the quadratic cost of softmax
attention by maintaining a recurrent fast-weight state, but it consistently lags on in-context
retrieval and long-context tasks. Existing remedies act on the write side of
memory through gating, delta updates, or kernel feature maps, but the read step is left unchanged: every past key contributes additively to the output, so useful targets are diluted by the bulk of stored vectors. We borrow one specific piece of softmax's geometry to construct a cheap read-time contraction of the query. A second-order Taylor expansion of the softmax log-partition at the isotropic-
attention point gives a local quadratic model whose
curvature coincides with the running key covariance, a quantity that can be maintained with the same recurrent/chunkwise mechanism as the linear-
attention state. The associated linear operator contracts the query along the high-variance directions of
memory before it reads the state. We call this mechanism
Curvature-Conditioned Query (CCQ). CCQ modifies only the read step and is composable with any linear-
attention backbone. Attached to GLA and Gated DeltaNet, it improves perplexity, zero-shot downstream accuracy, S-NIAH
retrieval at and beyond the training context, length-extrapolation perplexity from 4K to 20K, and LongBench accuracy.
Read the full abstract ▾
All authors Dong Le, Thong Nguyen, Cong-Duy Nguyen, Anh Tuan Luu
Computation and Language · Cross-listed from cs.SE
BenchEvolver: Frontier Task Synthesis via Solution-Centric Evolution
Yangzhen Wu, Aaron J. Li, Wenjie Ma et al. (+10 more)
The rapid progress of frontier large
language models has led to widespread
benchmark saturation, limiting the ability of existing
datasets to differentiate model capabilities or provide useful training signal. For instance, on LiveCodeBench, frontier models achieve over 99% Pass@1 on easy splits and exceed 90% Pass@1 on average across difficulty levels. Constructing new, challenging
datasets typically requires substantial human effort, creating a bottleneck for progress. We introduce BenchEvolver, a solution-centric evolutionary framework that automatically transforms existing coding problems into harder variants. Rather than generating problems from scratch, BenchEvolver evolves reference solutions through structured transformations and derives corresponding statements and tests from the evolved solutions. This design grounds generation in executable semantics, enabling scalable construction of high-quality, diverse, and difficult tasks with verifiable correctness. Applying BenchEvolver to LiveCodeBench and SciCode, we obtain evolved tasks that are substantially harder while maintaining validity, reference correctness, and diversity. We further curate LiveCodeBench-Plus, a 91-problem
benchmark combining evolved and difficult original LCB-v6 tasks, where frontier-model Pass@1 ranges from 27.5% to 62.6%, restoring clear discrimination among strong coding models. Importantly, evolved tasks remain challenging even for the model that generates them, enabling self-improvement. We further show that RL on evolved LCB tasks improves held-out coding performance: for gpt-oss-20b, seed+evolved training achieves +8.7 and +8.3 Pass@1 gains on LCB v6 Hard and LCB-Pro Easy, exceeding seed-only gains by 70.7% and 34.8%, respectively. Our results show that BenchEvolver can convert saturated
benchmarks into frontier-level evaluation suites and reusable training signal.
Read the full abstract ▾
All authors Yangzhen Wu, Aaron J. Li, Wenjie Ma, Li Cao, Ziheng Zhou, Mert Cemri, Shu Liu, Yuran Xiu, Chenxiao Yan, Haikun Zhao, Bin Yu, Ion Stoica, Dawn Song
Computation and Language
Worlds Within Words: Translating Culture in Ancient Chinese Texts with Multi-Agent Coordination
Xiaoqi He, Kaixin Lan, Mu You et al. (+3 more)
Large
language model (LLM)-based machine translation has advanced cross-cultural communication, yet it still struggles with culture-loaded words (CLWs) in ancient Chinese texts. The challenge extends beyond lexical
alignment to deciding when and how culture-dependent knowledge should be explicated for readers lacking relevant background. Literal translation often preserves surface forms while missing underlying concepts, whereas over-explicitation harms conciseness and readability. To address this problem, we formulate CLW translation as a selective explicitation task and propose MACAT, a Multi-
Agent Culture-Aware Translation framework that dynamically identifies culturally salient phrases and injects concise explanatory knowledge when necessary. MACAT further incorporates a quality-aware reranking module for candidate selection and a multi-round evaluation
agent that assesses translations across terminological precision, readability, fidelity, cultural preservation, and cultural explicitation. Experiments on traditional Chinese medicine (TCM) classics and the Analects show that, under a unified GPT-5.4 evaluation setting, MACAT consistently outperforms both the backbone model and general-purpose MT baselines on 100 TCM documents and a 20-chapter subset of the Analects.
Read the full abstract ▾
All authors Xiaoqi He, Kaixin Lan, Mu You, Tao Fang, Lidia S. Chao, Derek F. Wong
Computation and Language
IndoBias: A Dual Track Culturally Grounded Benchmark for LLMs Bias Evaluation in Indonesian Languages
Ikhlasul Akmal Hanif, Muhammad Falensi Azmi, Filbert Aurelian Tjiaranata et al. (+2 more)
Despite being home to more than 1300 ethnic groups and 700 indigenous
languages, bias in Large
Language Models has not been fully studied in Indonesia, thus leaving a critical gap in evaluating representational
fairness and localized stereotypes within its uniquely vast,
multilingual, and diverse sociocultural landscape. To address this, we introduce IndoBias as a culturally-grounded bias
benchmark to assess LLMs bias in Indonesian and three local
languages: Javanese, Sundanese, and Makasar. IndoBias features dual perspective evaluation tracks: depth-oriented (with contrastive-pairs) and breadth-oriented (with generation-based), where the latter is grounded in social science frameworks (SPI, O*NET, and WGI). Our results show that existing LLMs – particularly decoder models – exhibit strong bias towards prototypical sentences in Indonesian, while local
languages suffer higher bias under Ideology and Religion category. We also find that LLMs responses exhibit a non-uniform Stereotype Polarity when prompted with various local entities. Finally, we discover that, in Indonesian, Common Crawl texts introduce more bias during pretraining, compared to human-reviewed article texts (e.g., Wikipedia, News), whereas introducing local
languages to pretraining generally increases bias. This work highlights the importance of studying bias in culture-specific context. Warning: This paper contains example data that may be offensive, harmful, or biased.
Read the full abstract ▾
All authors Ikhlasul Akmal Hanif, Muhammad Falensi Azmi, Filbert Aurelian Tjiaranata, Eryawan Presma Yulianrifat, Fajri Koto
Computation and Language · Cross-listed from cs.LG
Beyond Sinusoids: A Morlet Wavelet Framework for Transformer Positional Encoding
Athanasios Zeris
Standard positional encodings for
transformers - sinusoidal and rotary (RoPE) - treat every position as equally local: they encode where a
token is, but not how far its positional influence should extend. We propose that the Morlet wavelet, which simultaneously minimises uncertainty in position and frequency, is the natural basis for positional encoding, and introduce Morlet Positional Encoding (MoPE): each
embedding dimension learns its own frequency and locality bandwidth from data. The main theoretical result is a unification: sinusoidal PE and the RoPE correlation kernel both emerge as limiting cases of MoPE when locality is switched off (sigma_i -> infinity). The phase of MoPE recovers the RoPE rotation angle exactly; the amplitude adds a learned Gaussian locality kernel that standard encodings lack. Empirically, MoPE combined with Energy-Gated
Attention achieves +0.119 improvement over standard
attention on TinyShakespeare, outperforming either component alone. Analysis of the learned parameters reveals that all 128 frequency-bandwidth pairs converge to the wavelet admissibility boundary - an empirical observation consistent with a companion result on energy gating, suggesting a reproducible property of character-level
language signals that warrants further investigation.
Read the full abstract ▾
Computation and Language
Agentic Clustering: Controllable Text Taxonomies via Multi-Agent Refinement
Simon Löwe and Emily Silcock
Recent text-clustering methods use large
language models to propose a cluster taxonomy from a corpus and then assign each text to it. These pipelines are fundamentally programmatic: the sequence of LLM calls and the rules for stopping, merging, and splitting clusters are fixed in code in advance, so they generalise poorly across corpora of different structure and cannot easily incorporate user-supplied constraints such as a target cluster count or a clustering intent. We propose an agentic alternative in which an orchestrator LLM inspects the state of the discovery process at each step and dispatches one of a small set of specialised
agents – proposer, synthesizer, auditor, investigator, and critic – adapting the pipeline to the corpus rather than executing a fixed one. On seven public text-clustering
benchmarks the method achieves state-of-the-art performance, beating the strongest prior baseline by 47% in mean ARI.
Read the full abstract ▾
Computation and Language
Understanding LLM Behavior in Multi-Target Cross-Lingual Summarization
Sangwon Ryu, Yihong Liu, Mingyang Wang et al. (+4 more)
Multi-target cross-lingual text summarization (MTXLS), which summarizes a source document into multiple target
languages, is increasingly important as users consume content in diverse
languages, but remains underexplored. To address this gap, we introduce multi-target cross-lingual element-aware (MEA), a new MTXLS
benchmark covering 24 target
languages. We
benchmark end-to-end and pipeline approaches across various LLMs and show that MTXLS performance still substantially lags behind English monolingual summarization. To better understand MTXLS in LLMs, we propose a layer-wise analysis framework for investigating how LLMs internally perform MTXLS. Our analyses suggest that translation and summarization behaviors emerge jointly within later layers rather than as distinctly decomposed stages. Most task-relevant processing occurs within these layers, and errors also tend to arise at similar depths. Motivated by these findings, we introduce an inference-time activation steering method that leverages hidden representations from English summarization to guide MTXLS generation. Experiments show that our method consistently improves MTXLS quality across target
languages.
Read the full abstract ▾
All authors Sangwon Ryu, Yihong Liu, Mingyang Wang, Yunsu Kim, Jungseul Ok, Gary Geunbae Lee, Hinrich Schuetze
Computation and Language · Cross-listed from cs.LG
Xingrun Xing, Haoqing Wang, Boyan Gao et al. (+2 more)
On-Policy
Distillation (OPD) is a fundamental technique for efficient post-training of large
language models (LLMs), with broad applications in
agent learning, multi-task enhancement, and model compression. However, OPD training becomes unstable when the teacher and student distributions differ substantially, as teacher supervision on student-generated
tokens may yield unreliable policy gradients and even cause optimization failure. This work addresses reliable on-policy
token-level supervision through credit assignment strategies, and proposes Trust Region On-Policy
Distillation, TrOPD. It features the following characteristics: 1) Trust-Region On-Policy Learning: TrOPD performs OPD only in regions where the teacher provides reliable supervision, mitigating the optimization difficulty of the K1 reverse-KL
estimator under distribution mismatch. 2) Outlier Estimation: For outlier regions, we explore gradient clipping, masking, and forward-KL estimation to reduce the adverse effects of unreliable supervision. 3) Off-Policy Guidance: The student continues generation from teacher prefixes and uses forward KL to imitate off-policy guidance, encouraging on-policy exploration toward reliable regions. Experiments show that TrOPD consistently outperforms SoTA OPD baselines, including OPD, EOPD, and REOPOLD, across mathematical
reasoning, code generation, and general-domain
benchmarks.
Read the full abstract ▾
All authors Xingrun Xing, Haoqing Wang, Boyan Gao, Ziheng Li, Yehui Tang
Computation and Language
Unlocking the Black Box of Latent Reasoning: An Interpretability-Guided Approach to Intervention
Shuochen Chang, Tong Bai, Xiaofeng Zhang et al. (+5 more)
Latent
reasoning enables Large
Language Models (LLMs) to perform multi-step inference within continuous hidden states, offering efficiency gains over explicit Chain-of-Thought (CoT). However, the opacity of these continuous thought vectors hinders their reliability and controllability. This paper bridges the gap between mechanistic interpretability and actionable control. We first present a systematic analysis using structural, causal, and
geometric probes, revealing that latent vectors encode compressed, faithful representations of
reasoning steps, with early vectors acting as critical causal hubs. Building on this, we operationalize these interpretability insights into a suite of training-free, decode-time interventions that refine the latent
reasoning process by imposing the identified
geometric and semantic priors. Extensive experiments across multiple model scales and diverse task domains demonstrate that our approaches consistently improve
reasoning accuracy. Our interpretability-guided interventions consistently unlock latent capabilities and improve
reasoning accuracy without any parameter updates.
Read the full abstract ▾
All authors Shuochen Chang, Tong Bai, Xiaofeng Zhang, Qianli Ma, Qingyang Liu, Zhaohe Liao, Yibo Miao, Li Niu
Computation and Language
Efficient RAG with Intent-Aware Retrieval and Semantics-Preserving Chunking
Fachrina Dewi Puspitasari, Chaoning Zhang, Jiaquan Zhang et al. (+6 more)
The demand for powerful instruction following and
reasoning capability of large
language models (LLMs) has promoted rapid development of
retrieval-augmented generation (RAG). The RAG system assists LLM generation by retrieving chunks of query-fit supplementary knowledge from an external database. Conventional RAG systems, however, suffer from information insufficiency due to two factors, which are intent-agnostic
retrieval and information fragmentation. Our work proposes a RAG framework, termed InSemRAG, that addresses these challenges via an iterative retrieve-and-check mechanism with two supporting modules, an intention-aware retriever (IAR) and semantics-preserving chunking (SPC). IAR implements a dynamic hybrid
retrieval method that adaptively weights the
retrieval channels based on the query intent, while SPC performs detection and reparation to the damaged evidence chunks to preserve the semantic integrity. To alleviate the computational latency brought by our iterative mechanism, we leverage small
language models (SLMs). Extensive experiments across several
benchmark datasets consistently demonstrate the competitiveness of our method against recent state-of-the-art RAG mechanisms. Particularly, our method achieves significant gains on multi-hop and evidence-sensitive tasks, with a 2.65-point improvement in F1 on HotPotQA and a 1.5-point increase in accuracy on FEVER. Our method also achieves competitive performance to Multi-Hop RAG with 4.32$\times$ lower latency with the utilization of SLM.
Read the full abstract ▾
All authors Fachrina Dewi Puspitasari, Chaoning Zhang, Jiaquan Zhang, Zhicheng Wang, Hafiz Shakeel Ahmad Awan, Rizwan Qureshi, Jewon Lee, Tae-Ho Kim, Yang Yang
Computation and Language
Connecting the Dots: Benchmarking Reflective Memory in Long-Horizon Dialogue
Jingjie Lin, Bingbing Wang, Zihan Wang et al. (+4 more)
Despite substantial progress in long-context modeling, existing
benchmarks remain confined to factual
memory for explicit recall, failing to measure the reflective
memory required to synthesize fragmented,
multimodal cues into high-level interpretations. To address this gap, we introduce RefMem-Bench, a
benchmark for reflective
memory in long-horizon dialogue. RefMem-Bench contains 26K annotated QA instances with eight reflective-
memory dimensions and three task formats, requiring models to move beyond surface-level
retrieval and infer latent meanings from evidence distributed across interaction histories. To enhance reflective
memory capability, we propose REflective
Memory INDuction (REMIND), a hierarchical framework that treats reflective
memory as progressive meaning construction. REMIND couples question-conditioned evidence
retrieval, salience-aware grounding, and abstraction-level supervision, and uses Progressive Reflective
Alignment to distill high-level reflective
reasoning into the factual inference pathway. Experiments show RefMem-Bench poses a substantial challenge to current models, while REMIND consistently improves both answer accuracy and
memory recall through progressive evidence perception, grounding, and abstraction.
Read the full abstract ▾
All authors Jingjie Lin, Bingbing Wang, Zihan Wang, Zhengda Jin, Weiming Qiao, Jing Li, Ruifeng Xu
Computation and Language · Cross-listed from cs.CV
Distilling Neuro-Symbolic Programs into 3D Multi-modal LLMs
Wentao Mo and Yang Liu
Current 3D spatial
reasoning methods face a fundamental trade-off: neuro-symbolic 3D (NS3D) concept learners achieve interpretable
reasoning through compositional programs but are constrained to closed-set concept vocabularies and simple programs; end-to-end 3D multi-modal LLMs (3D MLLMs) could handle complex natural
language and open-vocabulary concepts but suffer from black-box
reasoning without explicit spatial verification. We introduce APEIRIA, a neuro-symbolic 3D MLLM to bridge two paradigms by distilling symbolic
reasoning patterns into MLLMs with natural
language chain-of-thought. Our three-stage curriculum progressively builds
reasoning capabilities: a) 3D perception
alignment grounds object visual-
geometric features to the LLM, b) CoT-SFT teaches query decomposition and stepwise verification from symbolic program traces, and c) CoT-RL extends
reasoning patterns to open-set concepts and deeply nested instructions. By transferring
reasoning patterns rather than concept-specific knowledge, APEIRIA preserves key NS3D virtues: transparent
reasoning and modular interchangeability of planning and perception components. Evaluations on grounding, question answering, and captioning show that APEIRIA surpasses prior NS3D methods and matches state-of-the-art 3D MLLMs on 3D spatial
reasoning datasets, unifying symbolic methods' systematic
reasoning with MLLMs' flexibility. Code is available at https://github.com/oceanflowlab/APEIRIA.
Read the full abstract ▾
Computation and Language · Cross-listed from cs.CV
TECCI: Tricky Edits of Collected and Curated Images
Aishwarya Agrawal, Roy Hirsch, Yasumasa Onoe et al. (+2 more)
Despite tremendous recent progress, current text-guided image editing methods still struggle with many aspects of editing involving instruction following, minimally editing the source image, and ensuring high visual quality. These problems are especially apparent when the requested edit is challenging, such as those that involve position, motion, viewpoint, scale and creative edits. To systematically test generative image editors, we propose a novel image editing
benchmark – TECCI: Tricky Edits of Collected and Curated Images. TECCI consists of a completely new set of images we are releasing. The images in TECCI span 7 image categories. The images and these categories were curated intentionally to target weaknesses of existing methods. The edit instructions in TECCI are automatically generated by Gemini, covering 5 edit types per source image. We also curated a set of 530 images for which we created challenging manually written edit instructions. Overall, TECCI contains 7550 pairs of images and edit instructions. We conduct human evaluations of five leading image editing models on TECCI. Humans judge outputs along three dimensions: 1) instruction following, 2) minimality of the edits, and 3) visual quality. To scale-up the evaluation, we also build an auto-rater using Gemini that achieves 74.7% accuracy in matching human evaluations. Our evaluations reveal that: 1) none of the models exceed a 22% overall success rate, demonstrating the challenging nature of TECCI, 2) Nano Banana Pro is the best performing model overall, 3) models perform significantly better at instruction following compared to minimal edits and visual quality, 4) models struggle with editing architecture and nature images which require strong understanding of spatial layout and intricate visual details. 5)
reasoning and creative edits are the most difficult, whereas color and appearance edits are the easiest.
Read the full abstract ▾
All authors Aishwarya Agrawal, Roy Hirsch, Yasumasa Onoe, Sherry Ben, Jason Baldridge
Computation and Language
DiscourseFlip: An Oblique Discourse-Level Opinion Manipulation Attack against Black-box Retrieval-Augmented Generation
Yuyang Gong, Miaokun Chen, Jiawei Liu et al. (+5 more)
Retrieval-Augmented Generation (RAG) systems are widely deployed and increasingly influential, but their reliance on external corpora exposes new security risks from poisoned
retrieval content. Existing RAG attacks are largely focusing on individual queries or narrow topic-local query sets, which limits their practical reach and offers limited camouflage in real-world settings. In this paper, we introduce discourse-level opinion manipulation, a new threat model in which coordinated influence across a semantic query network induces opinion shifts over a holistic, multi-topic query space. We formalize this threat in a black-box setting and propose DiscourseFlip, an agentic, graph-guided attack that dynamically allocates a limited poisoning budget to maximize discourse-level opinion deviation. Extensive experiments demonstrate that DiscourseFlip consistently induces targeted opinion shifts across the contextualized query network and significantly outperforms existing baselines in terms of coverage and effectiveness. User studies further confirm that DiscourseFlip is effective while remaining well camouflaged from user detection. Moreover, systematic analyses show that existing mitigation strategies are ineffective against discourse-level manipulation, underscoring the urgent need for more robust and adaptive defenses to address discourse-level vulnerabilities.
Read the full abstract ▾
All authors Yuyang Gong, Miaokun Chen, Jiawei Liu, Zhuo Chen, Guoxiu He, Wei Lu, XiaoFeng Wang, Xiaozhong Liu
Computation and Language
Implicit Geographic Inference in LLM Medical Triage: Language-Driven Disparities in Emergency Recommendations
Qi Han Wong
We investigate whether large
language models produce different medical triage recommendations for identical symptoms based solely on the
language of the patient
prompt. Using Gemini 3.5 Flash, we evaluate a neurological symptom profile (persistent headache, blurred
vision, nausea) across six
languages (English, Spanish, Chinese, Hindi, Japanese, Arabic) with 30 runs per condition (n=450 total API calls). We find that the model recommends emergency room visits at rates ranging from 0% (Japanese, Hindi) to 30% (English, Arabic), despite assigning nearly identical severity scores (7.7-8.0/10) across all
languages. Adding a single sentence specifying the patient's US location increases ER recommendations by up to 76.7 percentage points for non-English
prompts, while the reverse anchor (English
prompt with a Tokyo location) reduces the ER rate from 30% to 6.7%. A back-translation control (Japanese to English) produces ER rates comparable to the English baseline, confirming that the disparity is not caused by translation quality but by implicit geographic inference from the input
language. We release the complete
dataset, experiment code, and results.
Read the full abstract ▾
Computation and Language · Cross-listed from cs.AI
The Shape of Wisdom: How Language Models Converge on an Answer
Shailesh Rana
When a
language model answers a factual question, where does its internal state go? We asked four open models (Llama-3.1-8B, Qwen2.5-7B, Gemma-2-9B and Llama-3.2-1B) 2,256 factual questions whose answers recur, such as the country a person is a citizen of. At every layer, we recorded the state from which the answer is produced. Questions that receive the same answer converge on a shared region of state space, which we summarise by its centre, the answer centroid. The answer becomes readable within a single layer. In every model, the share of questions whose state lies nearest its own answer's centroid jumps at one layer, between 47% and 71% of the way through the network. The centroids follow the answer the model gives rather than the question: the centroid of a country learned from questions about people's citizenship also identifies that country in questions about towns and films. The model also uses the centroids. After the jump, moving the state of a wrongly answered question to the right answer's centroid makes the model give the right answer in 83%-97% of cases late in the network, against 2%-4% for a random move of the same size. Before the jump, moves of the same size almost never work. Near misses, in which the model answers wrongly but picks the right option in multiple choice, can be seen from the onset layer, where the right answer's centroid is already among the nearest to the state. This signal is about as informative as the right answer's rank in the model's final output, and more informative than the model's confidence in its reply.
Read the full abstract ▾
Computation and Language
Low-Resource Safety Failures Are Action Failures, Not Representation Failures
Rashad Aziz, Ikhlasul Akmal Hanif and Fajri Koto
Language models often answer harmful requests in low-resource
languages (LRLs) that they refuse in high-resource
languages (HRLs). Across three instruction-tuned models and 23
languages, harmful refusal falls from 87.9% in HRLs to 43.9% in LRLs, while harmless refusal remains low. A common explanation is that models represent harmfulness weakly in LRLs. We test whether harmfulness is instead represented but does not reliably produce refusal. Across three models, a harmfulness direction learned from HRL activations still separates harmful from harmless LRL
prompts, showing that complete absence of harmfulness information cannot explain many failures. However, harmfulness scores shift downward for LRL
prompts, making harmful
prompts less likely to reach the range associated with refusal. Motivated by this shift, we train a low-rank logistic classifier on HRL activations and calibrate its threshold with a few target-
language examples. During generation, the classifier conditionally adds or ablates the HRL harmfulness direction. With the same HRL data and 32 target-
language examples per class, CAST remains limited by low harmful refusal and AdaSteer by high harmless refusal, yielding mean refusal selectivity ($\Delta$ = harmful - harmless refusal) of 33.6 and 6.8, respectively. Our intervention reaches 54.5 while preserving MMLU utility. HRL-only calibration improves selectivity for Qwen and Gemma, whereas Llama benefits from target-
language calibration. These results show that recalibrating existing representations can offer a training-free method for repairing low-resource safety failures.
Read the full abstract ▾
Computation and Language
CA-BED: Conversation-Aware Bayesian Experimental Design
Daniel Arnould, Rashad Aziz, Zixuan Kang et al. (+5 more)
Large
Language Models (LLMs) excel at static
reasoning tasks, yet their performance often degrades in interactive scenarios where information must be actively acquired through questioning. A key challenge lies in selecting questions that reduce uncertainty while incorporating responses that may be ambiguous or only partially informative. To address this, we propose Conversation-Aware
Bayesian Experimental Design (CA-BED), an inference-time probabilistic dialog planning framework that integrates
Bayesian Experimental Design with LLM-based likelihood estimation to optimize question selection over multiple conversational turns. CA-BED maintains a belief distribution over hypotheses, anticipates possible answers, and propagates expected information gain through a simulated conversation tree. Across two structured entity-deduction
benchmarks, CA-BED yields an average 21.8% improvement in success rates over direct prompting, with comparable gains relative to alternative information-seeking methods. It achieves these gains with an average increase of only 1.8 conversational turns compared to direct prompting.
Read the full abstract ▾
All authors Daniel Arnould, Rashad Aziz, Zixuan Kang, Tanav Changal, Kevin Zhu, Sunishchal Dev, Gabriel Grand, Shreyas Sunil Kulkarni
Computation and Language
Thinking Economically: A Hierarchical Framework for Adaptive-Complexity Reasoning in LLMs
Yubo Gao, Haotian Wu, Hong Chen et al. (+8 more)
Chain-of-Thought (CoT) has significantly enhanced LLM
reasoning, yet often incurs substantial computational overhead due to "overthinking": generating excessively long rationales without commensurate accuracy gains. Existing efficiency methods typically apply uniform compression, which overlooks a critical observation that
reasoning complexity is heterogeneous at two distinct granularity: across different problems and within individual
reasoning steps. This motivates our principle of Thinking Economically: intelligently allocating computational resources based on intrinsic task and step demands rather than pursuing uniform brevity. We propose Hierarchical Adaptive Budgeter (HAB), a training framework that operationalizes this principle through coarse-to-fine budgeting. At the inter-step level, HAB predicts the optimal
reasoning depth for each problem. At the intra-step level, HAB learns step-specific
token budgeting signals from PPL-derived step comparisons and an adaptive Pareto optimization objective that captures the local quality-efficiency trade-off, while a Fisher Information-based pruner further provides fine-grained training-time guidance, thereby encouraging the generator to internalize more economical
reasoning patterns. Experiments on GSM8K and MATH500 show that HAB not only surpasses standard CoT in accuracy but also reduces
token usage, achieving a stronger performance-efficiency trade-off than the compared baselines.
Read the full abstract ▾
All authors Yubo Gao, Haotian Wu, Hong Chen, Junquan Huang, Yibo Yan, Jungang Li, Zihao Dongfang, Sicheng Tao, Puay Siew Tan, Jie Zhang, Xuming Hu
Computation and Language · Cross-listed from cs.CR
BraveGuard: From Open-World Threats to Safer Computer-Use Agents
Yunhao Feng, Xiaohu Du, Xinhao Deng et al. (+13 more)
Computer-use
agents extend
language models from text generation to sustained interaction with files, terminals, browsers, and external tools. This shift creates safety risks that are difficult to detect from isolated
prompts or final responses, because harm often emerges only through multi-step execution traces whose individual actions appear locally benign. We introduce BraveGuard, a self-evolving defense framework for training guard models from open-world threat signals and realistic
agent trajectories. BraveGuard mines recent research sources to identify emerging risks and attack patterns, instantiates them as executable computer-use tasks, collects
agent rollouts, and derives trajectory-level supervision for guard model training. As new threats and validation failures appear, the pipeline can be repeated, yielding an adaptive defense loop rather than a static,
benchmark-driven training process. We instantiate BraveGuard by training multiple guard backbones, including Qwen3-Guard and Llama-Guard variants, and evaluate the resulting guards on trajectory-level
agent-safety
benchmarks. BraveGuard consistently improves safety detection across computer-use trajectories. On AgentHazard, it substantially improves detection accuracy over off-the-shelf guard models, with accuracy increasing from 38.79% to 82.38% under the averaged guard-model setting. These results show that guard supervision grounded in open-world threat discovery and realistic
agent execution can improve safety monitoring beyond fixed taxonomies and synthetic
prompt-level data. BraveGuard offers a scalable path toward adaptive defenses for computer-use
agents facing evolving real-world risks.
Read the full abstract ▾
All authors Yunhao Feng, Xiaohu Du, Xinhao Deng, Yifan Ding, Ming Wen, Yixu Wang, Yuxiang Xie, Baihui Zheng, Yingshui Tan, Yige Li, Yutao Wu, Kerui Cao, Wenke Huang, Yanming Guo, Xingjun Ma, Yu-Gang Jiang
Computation and Language
Not All Explanations Simulate Equally: Comparing Verbalized Feature Attributions and Self-Generated Rationales
Pingjun Hong and Benjamin Roth
Natural-
language explanations are often treated as a unified interface for understanding model behavior, but different explanation sources may support simulation in different ways. This paper compares two families of explanations for question answering models: verbalized feature attributions and self-generated rationales. We evaluate them under a shared counterfactual simulation setting, using an LLM judge as predictor and measuring whether it can better predict a model's answers to follow-up questions when given its explanation. Across multiple instruction-tuned models, we analyze how explanation source, verbalization strategy, and feature granularity affect the simulatability of explanations. Our results show that explanation format and granularity affect simulatability: attribution-based explanations and self-generated rationales differ in how much they improve counterfactual prediction, with effects that vary across models and formats.
Read the full abstract ▾
Computation and Language
From Outliers to Errors: Auditing Pali-to-English LLM Translations with Multi-Reference Adjudication
Máté Metzger, Nadnapang Phophichit and Hansa Dhammahaso
Single-score translation metrics can conflate legitimate variation with error, a problem especially acute for classical
languages where multiple defensible English renderings of the same passage coexist. We audit Pali-to-English output from four flagship large
language models (LLMs): GPT-5.5, Claude Sonnet 4.6, Gemini 3.1 Pro, and Grok 4.3, on 1,700 passages from the Pali Canon, using three established human translations by Bhikkhu Sujato, Thanissaro Bhikkhu, and Bhikkhu Bodhi as a local reference envelope rather than a single gold standard. Each candidate's normalized
embedding drift from the reference centroid serves as a triage signal, not an error label; the 1,203 candidates above a 1.5 drift threshold are then adjudicated by a blinded three-model LLM judge panel, calibrated against a 300-instance author-adjudicated validation set. Two results stand out. First, drift predicts severity rather than error per se: the major-error rate among adjudicated high-drift candidates rose monotonically from 7.9% in the 1.5-2.0 band to 51.6% above 3.0, while approximately 80% of 1.5-2.0 outliers were judged valid translation variations. Second, model differences were clearest in the high-drift tail: GPT-5.5 had the lowest adjudicated high-drift major-error rate, with confidence intervals overlapping those of Claude Sonnet 4.6 and Gemini 3.1 Pro; Grok 4.3 had both the largest outlier volume and the highest tail major-error rate (27.6% overall, 74.4% above drift 3.0). The dominant major-error categories (e.g. omission or truncation, doctrinal term errors) are precisely the failures most likely to mislead readers of doctrinal text. The contribution is a reusable audit design for classical-to-modern translation: define a local reference envelope from multiple human translators, use
embedding drift to prioritize review, and adjudicate the flagged tail rather than treating outlier status as error.
Read the full abstract ▾
Computation and Language · Cross-listed from cs.DL
Digging Up Citations: FOSSIL, a Dataset and Workflow for Reference Extraction in Law and the Humanities
Luca Foppiano and Christian Boulanger
Citation extraction tools are designed for the structured end-of-document bibliographies of the natural sciences, but law and humanities scholarship cites references primarily in footnotes, where bibliographic data is interleaved with commentary and cross-references and varies widely across
languages and styles. To address the scarcity of suitable gold-standard resources, we present FOSSIL (Footnote-based Open-access SSH Scientific Instance Labels), an openly licensed
multilingual dataset of 96 annotated scholarly articles containing over 7,600 footnote-embedded references, together with PDF-TEI Editor (a collaborative web annotation tool), a documented seven-annotator workflow, and a Grobid specialization for footnote-based citations. In end-to-end evaluation, the specialized pipeline nearly doubles extraction quality over default Grobid (micro-F1 from 0.36 to 0.72), driven largely by improved recall, while showing that substantial headroom remains for cross-references and mixed-content footnotes. This extended abstract presents work in progress; annotations of citations
segmentation and parsing, and cross-reference resolution are ongoing.
Read the full abstract ▾
Computation and Language
MiCU: End-to-End Smart Home Command Understanding with Large Language Model
Haowei Han, Kexin Hu, Weiwei Cai et al. (+6 more)
Command understanding systems in smart home ecosystems can automate device control and substantially improve user experience. However, while they perform well on precise utterances (e.g., "turn on the bedroom light"), they struggle with ambiguous or misaligned commands (e.g., "make the bedroom cozy"). Large
language models (LLMs) generalize well across various domains and can outperform traditional rule-based systems on such tasks, but their effectiveness is often constrained by scarce domain-specific data, insufficient task-specific adaptation, and high computational costs. In this paper, we propose an automated training data synthesis workflow using user logs and LLMs; then we build MiCU, a domain-specific LLM that excels at command understanding. Specifically, we employ curriculum learning to inject domain knowledge into the base LLM, then we enhance its
reasoning ability via cold-start training combined with reinforcement learning (RL) guided by domain-specific thinking rules. Additionally, we introduce a
token compression technique that condenses device description into a single special
token, substantially reducing inference overhead and enabling \model-fast, an efficient variant optimized for long inputs. Extensive experiments show that MiCU significantly outperforms baselines, with an average accuracy gain of 20.01% across all device categories. We have deployed MiCU in the Xiaomi Home app, receiving approximately 1.7 million page views per day. Production evaluations show that MiCU reduces user correction rate by 1.57% and increases human audited accuracy by 32.05%. Our data and code are available at https://github.com/xiaomi-research/iot_spec_llm
Read the full abstract ▾
All authors Haowei Han, Kexin Hu, Weiwei Cai, Debiao Zhang, Bin Qin, Yuxiang Wang, Jiawei Jiang, Xiao Yan, Bo Du