Peer Revenue Publish or profit

Papers

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

Computer Science Computation and Language

cs.CL Watch Computation and Language

Thursday 6 August

Computation and Language · Cross-listed from cs.AI

DreamGuard: Efficient Runtime Guardrail for LLM Agents via Risk-Aware World Model

Wenhao Lin, Chenyu Yu, Xingwei Lin et al. (+6 more)

As large language model (LLM) agents increasingly invoke external tools and interact with real-world systems, unsafe actions may cause irreversible consequences on external states, user data, and downstream services. Recent runtime guardrails mitigate such risks by checking proposed actions before execution, but many remain reactive: they primarily assess the apparent safety of the current action, lacking an explicit model of how risk evolves across the trajectory. This limitation creates a critical blind spot for long-horizon risks, where individually benign-looking actions can gradually drift the agent toward hazardous states. In response, we propose DreamGuard, a proactive guardrail for LLM agents built around a risk-aware world model. The world model maintains a compact recurrent latent state over the trajectory and predicts future latent states from which DreamGuard derives immediate-hazard and prefix-risk evidence. It then fuses these multi-horizon signals into intervention decisions before execution. Experiments across four benchmarks and an online guardrail evaluation show that DreamGuard outperforms generic, reactive, and proactive guardrail baselines, achieves the best safety-utility trade-off among evaluated guardrails, and maintains an average end-to-end latency of 25 ms per call.
Read the full abstract

All authors Wenhao Lin, Chenyu Yu, Xingwei Lin, Sicong Cao, Xiang Chen, Lei Xue, Le Yu, Letian Sha, Chunming Wu

Computation and Language

Answer First, Reason Later: When Commitment Order Costs Accuracy in Diffusion Language Models

Jewon Yeom, Jaewon Sok, Seonghyeon Park et al. (+3 more)

Masked diffusion language models revise many masked output positions in parallel. We call a token committed once it becomes visible and is never masked again, and call a response answer-first when the final answer commits before the reasoning printed ahead of it. On 1,069 GSM8K test questions, an explicit step-by-step instruction increases the accuracy difference between unrestricted decoding and a decoder that permits commitment only near the left-most unresolved position; unrestricted decoding also produces more answer-first trajectories. On MATH-500, the two LLaDA models spend most of a short output canvas on reasoning that commits after the answer, and the benefit of frontier gating decreases as that postanswer writing disappears. Dream-7B has little post-answer writing and follows a different accuracy pattern. A controlled four-option task reserves a one-token answer position before generation. Delaying that position outperforms an equally timed reasoning-token delay on LLaDA-8B, LLaDA-1.5, and Dream-7B. The raw difference is largest on Dream, whose free accuracy on the controlled task is lower. Answers commit much earlier under the reserved-position interface than in ordinary free-form generation, which limits how far the intervention result can be generalized. Commitment order affects the context used to complete a response and the allocation of a finite output canvas.
Read the full abstract

All authors Jewon Yeom, Jaewon Sok, Seonghyeon Park, Jeongjae Park, Hwiyeong Lee, Taesup Kim

Computation and Language · Cross-listed from cs.LG

Reasoning Errors Have a Region and a Direction in the Residual-Stream Trajectory of LLMs

Hamed Damirchi, Ignacio Meza De la Jara, Damith Ranasinghe et al. (+2 more)

As language models are increasingly used for tasks that require verifiable reasoning, reliably distinguishing sound reasoning from flawed reasoning has become an important practical problem. Recent trajectory-based methods seek this signal in layerwise residual-stream displacements, which capture how representations change while attenuating some stable, token-specific information. However, displacement omits the state from which an update originates, whereas restoring the full state risks reintroducing shortcut-prone information. We identify this trade-off and propose a three-stream detector that combines motion with two restricted views of location. A coarse region reader based on vector quantization and a fine direction reader over normalized multi-layer states. This design restores enough state context to interpret the motion without returning to full-state probing. On reasoning benchmarks unseen during training, our method improves selection accuracy by up to 12% over the displacement-only state of the art and 21% over single-layer probing baselines. Although trained only on reasoning benchmarks, it also reads factual completion and fact verification, ahead of every detector we compare against, which places the signal on correctness rather than on a kind of reasoning. Ablations further show that motion, region, and direction provide complementary signals. These results suggest that reasoning validity is better read from state-conditioned motion than from either static states or decontextualized trajectories alone.
Read the full abstract

All authors Hamed Damirchi, Ignacio Meza De la Jara, Damith Ranasinghe, Yuhang Liu, Javen Shi

Computation and Language

Relay, Don't Route: Adaptive Population Handoff for Cost-Efficient LLM-Driven Evolution

Sichun Luo, Yi Huang, Guanzhi Deng et al. (+6 more)

Large language model (LLM)-driven evolution has shown promise for program search and algorithm discovery, but relying on strong models throughout long evolutionary runs is costly. A natural alternative is to combine cheap and strong models under a fixed inference budget. However, existing approaches typically allocate models at the level of individual queries or mutation steps, overlooking that evolutionary search is stateful: each generated candidate changes the population from which subsequent mutations are produced. We empirically analyze LLM-driven evolutionary trajectories and find that search progress is strongly front-loaded, early trajectory performance is informative but noisy, and cheap models recover much of the early progress achieved by strong models at lower cost. Motivated by these findings, we propose \model, a training-free framework that shifts budget allocation from individual calls to evolving populations through adaptive population handoff. A cheap model explores multiple trajectories in short blocks allocated by a bandit scheduler. Relay Gain, defined as the marginal improvement of a compact, quality-diverse candidate bank constructed for handoff, serves as the scheduler reward and determines when to hand off. The curated candidates initialize a shared strong model population for refinement. Across four benchmarks and three budgets, \model achieves the highest mean score in 11 of 12 settings, outperforming competitive baselines. Our results suggest that in stateful search, budget allocation should be organized around the population, not the individual call.
Read the full abstract

All authors Sichun Luo, Yi Huang, Guanzhi Deng, Haibo Wang, Haochen Luo, Lei Li, Zefa Hu, Junlan Feng, Qi Liu

Computation and Language · Cross-listed from cs.AI

Refining Over Resampling: Test-Time Self-Correction for LLM Reasoning

Ahsan Bilal, Muhammad Ahmed Mohsin, Muhammad Umer et al. (+4 more)

Test-time scaling improves LLM reasoning by using additional inference compute, but wider sampling alone can suffer from diminishing returns: new rollouts often repeat existing answer patterns instead of adding useful reasoning diversity. Verifier-based selection offers an alternative, but its performance depends on the calibration of an external reward model. We propose a verifier-free breadth–depth refinement framework that uses test-time compute to both explore and improve candidate solutions. The method samples multiple independent reasoning rollouts, refines each rollout through iterative self-critique and self-correction, and aggregates the refined answers by majority voting. Breadth preserves diverse initial attempts, while depth repairs local reasoning errors before aggregation. Across AIME24, AIME25, AMC, OlympiadBench, and MATH500, our method consistently improves over greedy decoding, majority voting, verifier-based best-of-$N$, beam search, and lookahead decoding across multiple open-weight models. For instance, with Qwen2.5-1.5B, accuracy increases from the strongest verifier-based baseline to $58.0\%$ on MATH500, and from $25.0\%$ to $32.5\%$ on AMC. These results show that test-time compute can be more effective when used to refine sampled trajectories rather than only to sample more candidates or rely on verifier-guided selection.
Read the full abstract

All authors Ahsan Bilal, Muhammad Ahmed Mohsin, Muhammad Umer, Lena Trigg, Ali Subhan, Muhammad Ali, Dean F. Hougen

Computation and Language

Human-Like Anaphor Resolution in Large Language Models

Keane Zhang, Varshini Chinta, Raj Sanjay Shah et al. (+1 more)

Anaphors are expressions that refer to other expressions, called antecedents. The process of connecting the two is called resolution. Cognitive science has identified multiple factors that affect the speed and success of anaphor resolution, including discourse structure, situation-model properties, and semantic factors. Here, we investigate whether these factors also affect anaphor resolution in five Large Language Models (LLMs) with open weights: GPT-2-XL, Llama-3.1-8B, Pythia-12B, Mistral-7B, and Mistral-24B. To model processing difficulty, we adopt the standard linking hypothesis that relates human reading times to model surprisal at the anaphor. As a second behavioral measure, we compare model accuracy to human accuracy on comprehension questions probing the antecedents of anaphors. The results show selective cognitive alignment: some LLMs exhibit human-like sensitivity to discourse prominence and distance-based factors in anaphor resolution, while showing weaker or absent sensitivity to semantic interference effects. These findings delimit the conditions under which LLMs approximate human anaphor resolution.
Read the full abstract

All authors Keane Zhang, Varshini Chinta, Raj Sanjay Shah, Sashank Varma

Computation and Language · Cross-listed from cs.AI

Measuring and Detecting Harmful AI Sycophancy

Bohan Jiang, Dawei Li, Yasin Silva et al. (+1 more)

Sycophantic responses are becoming pervasive in large language models (LLMs), and prior work has pointed out that some of them could be harmful. This paper focuses on one harmful sycophancy: preference-induced stance reversal sycophancy (PSRS), where a model reverses an initial stance merely to align with a user's stated preference. While existing research mainly measures how sycophantic a model is, we go further and ask whether PSRS can also be detected automatically from a single response. To investigate this at scale, we introduce CAP (Contrastive Anchor Probing), a framework for collecting labeled PSRS data. Applying CAP to 17 open- and closed-source LLMs, we collect 290,460 labeled responses across 12 everyday-advice domains. We organize our study around three research questions. (1) How often does PSRS occur? (2) How well can it be detected? (3) How does detection generalize to unseen models? We first reveal that PSRS rates range from 5% to 56% across LLMs, with more capable models being less sycophantic. Next, we show that detecting PSRS is feasible from the response text alone, and detectors need to learn subtle PSRS patterns from the training data. Because new LLMs appear rapidly, detectors inevitably encounter unseen models, making cross-model generalization an important framework goal. We demonstrate that detection performance drops on unseen models and propose an initial approach to address this challenge. We will release our dataset and code to support future research.
Read the full abstract

All authors Bohan Jiang, Dawei Li, Yasin Silva, Huan Liu

Computation and Language

FOCUS: Decoupling Expert Personas in LLMs to Enhance Domain Expert Capabilities

Guanyu Wang, Zidi Zhang and Xu Chu

Large Language Models (LLMs) can exhibit diverse personas, and activating expert personas has been shown to improve domain expertise and task accuracy. However, existing persona control methods often suffer from cross-domain coupling, which may lead to overly aggressive behavior in high-caution domains such as healthcare, or excessive conservatism in risk-sensitive domains such as financial trading. To address this issue, we propose FOCUS (\textbf{\underline{F}}ine-tuning with \textbf{\underline{O}}rthogonal \textbf{\underline{C}}ontrol for \textbf{\underline{U}}ncoupled persona\textbf{\underline{S}}). FOCUS first automatically extracts expert persona vectors from LLMs, then applies orthogonal decomposition to decouple domain-specific expert personas, and finally introduces an expert gating module to adaptively control persona activation according to task contexts. With a two-stage training strategy and a gated selection regularizer, the model learns to activate appropriate personas for both single-domain and cross-domain tasks. Experiments on financial, legal, medical, and cross-domain benchmarks show that FOCUS improves task accuracy and outperforms existing persona control methods. Our code is available at \href{https://anonymous.4open.science/r/openpersona-48F4}{this url}.
Read the full abstract

Computation and Language

SkillZip: Contract-Preserving Graph Compression for Scalable Agent Skill Libraries

Xingyu Tan, Xiaoyang Wang, Qing Liu et al. (+4 more)

Large Language Models (LLMs) increasingly act as agents whose procedural knowledge is stored in reusable skill packages and loaded at inference time. As skill libraries grow, a central challenge is to expose the smallest sufficient executable context under a limited context budget. Existing systems struggle to reuse routines below the whole-skill level, preserve procedural contracts during compression, keep compressed routines executable and expandable, and update the compressed library as skills evolve. These challenges reveal a unit mismatch: skills are retrieved as packages, compressed as text, and converted into execution graphs only after retrieval, whereas reliable reuse requires a contract-bearing procedural unit. We propose SkillZip, an execution-aware procedural abstraction framework that performs contract-preserving compression over section-level graphs. SkillZip rewrites recurring contract-valid motifs into reversible ported macros while preserving boundary signatures, dependency closure, verifier reachability, and source-level expansion. At inference time, it hydrates a compact, dependency-closed context and expands macros only when required. ReZip further integrates new skills and revises risky macros using execution evidence. Comprehensive experiments1 on technical and embodied agent benchmarks show SkillZip consistently outperforms the strongest baseline by up to 12.2 points, while achieving a 3.46x compression ratio with 99.2% dependency preservation and 98.7% verifier reachability. Scaling analyses further confirm robust retrieval across skill libraries ranging from 200 to 100K skills.
Read the full abstract

All authors Xingyu Tan, Xiaoyang Wang, Qing Liu, Xiwei Xu, Xin Yuan, Liming Zhu, Wenjie Zhang

Computation and Language

Where Models Converge and Humans Diverge: A Coverage Framework for Distributional Pluralism in Open-Ended Generation

Zini Yang, Emily Wenger and Richard So

When a large language model (LLM) writes Harry Potter fanfiction, it reliably produces fundamental elements of the Hogwarts universe, such as recognizable places and characters. Human-written Harry Potter fanfictions, however, typically include these fundamentals and much more, incorporating stylistically irregular content and relationship-diverse plotlines. This gap between LLM and human writing has been noted across a variety of domains. LLMs tend to produce "average" writing, while human writing contains more diverse content that covers a broader distribution. Existing work has shown the existence of this distributional "gap", but no work has proposed a systematic way to measure it. Our paper proposes a human-grounded framework that uses the empirical distribution of human writing on a topic to measure the distributional breadth of LLM-generated content on that same topic. We propose two metrics, LLM Coverage (LLM-Cov) and In-Boundary Rate (IBR), that separate the plausibility of LLM content from its distributional breadth. Across ideation and narrative tasks, we find that current LLMs produce plausible but narrow content that concentrates near the center of the human response space. Our framework can enable researchers to better assess the distributional breadth of LLM-authored content, which we term its "cultural reach".
Read the full abstract

Computation and Language · Cross-listed from cs.CV

From Sports to Safety: Benchmarking Proactive Risk Inference in MLLMs

Jiawei Qiu, Yichen Xu, Jianzhe Ma et al. (+5 more)

Timely anticipation of physical hazards is essential for real-world safety, yet existing MLLM evaluations focus on harmful content or general risks, leaving proactive physical hazard prediction underexplored. Sports provide a well-suited testbed: accident causes span diverse injury dimensions and pre-accident spatiotemporal cues draw on reasoning capabilities shared with broader safety domains such as autonomous driving and fall detection. We introduce SPRINT (Sports Proactive Risk INference Testbed), a benchmark of 2,888 real-world sports videos (2,440 accident, 448 safe controls) spanning 14 sports and 3 environmental settings. Accident videos feature fine-grained annotations of early hazard cues, accident timing, and hierarchical causes; safe videos are manually verified as accident-free and serve to diagnose prompt-induced false alarms. Evaluating state-of-the-art MLLMs under diverse prompts and temporal windows reveals a sharp gap between hazard sensitivity and understanding: the best model exceeds 95% in signaling hazards yet falls below 50% in identifying their causes. Diagnostic experiments further show that explicit danger queries trigger severe false alarms even on hazard-free videos. These findings indicate that current MLLMs exhibit only superficial proactive safety, lacking stable, cause-grounded early warning, and underscore the need for reliable proactive safety in dynamic physical environments. Data and code will be open-sourced upon acceptance.
Read the full abstract

All authors Jiawei Qiu, Yichen Xu, Jianzhe Ma, Mingyang Yu, Wenbin Zhu, Yang Han, Pinzheng Lv, Wenxuan Wang

Computation and Language · Cross-listed from cs.AI

EcoAgent-Bench: Evaluating Economic Decision-Making in Budget-Constrained LLM Agents

Jie Wu, Ming Gong, Feixiang Cheng et al. (+1 more)

Agent benchmarks usually measure task completion and treat resource use as an auxiliary statistic. In deployment, however, the choice among a local lookup, broad search, composite research tool, stronger model, or human escalation is part of the task itself. We introduce EcoAgent-Bench, in which every task specifies priced actions and an explicit budget. Its 304 real-derived tasks span five families adapted from GAIA, HotpotQA, and MuSiQue, and test four decisions: avoiding unnecessary escalation, escalating when local evidence is insufficient, selecting a model tier, and stopping on unsupported premises. We evaluate seven LLM agents in tool-API and workspace-CLI settings, together with four oracle scripted controls. Micro-averaged accuracy rewards one-sided policies: always-escalate controls achieve high micro success while failing save-oriented tasks. We therefore also report an economic-consistency score (the worse of accuracy on upgrade-oriented and save-oriented family groups) which exposes this failure. Tool-API agents attain only 3.9-24.0% micro strict success (at most 7.3% economic consistency), often either stopping before warranted escalation or overspending on cheap tasks. A threshold-crossing budget sweep changes GPT-5.4's escalation rate from 0% to only 3%. These results show that completion under a budget and economical action selection are distinct properties. We release the task bundle, transformation pipeline, frozen evaluation environments, and integrity-bound result artifacts needed to study both.
Read the full abstract

All authors Jie Wu, Ming Gong, Feixiang Cheng, Qinqin Zhao

Computation and Language

Different Perturbations, Different Mechanisms: Understanding Continued Pre-training for Zero-Shot Dialect Robustness

Aarohi Srivastava and David Chiang

Dialectal variation remains a major challenge for multilingual language models. Perturbation-based continued pre-training (CPT) has emerged as a promising approach to improving robustness, yet existing work largely evaluates individual perturbation strategies in isolation and provides limited insight into why they work. We present a systematic study of perturbation-based CPT for multilingual dialect robustness in LLMs, comparing six training conditions across nine German, Italian, and Arabic dialect tasks. Perturbation-based CPT, especially character-noised CPT, consistently improves zero-shot dialect robustness while largely preserving standard variety performance. More importantly, we show that methods with similar downstream performance induce distinct mechanisms of robustness, exhibiting different patterns of language model adaptation, representational alignment, and prediction repair. Our results provide a more complete understanding of how synthetic surface variation improves robustness and offer practical guidance for selecting CPT strategies in multilingual and dialectal settings.
Read the full abstract

Computation and Language · Cross-listed from cs.PL

Learning Context-Free Grammars for Grammar-Constrained Decoding via Declarative Agentic Programming with Guarantees

Kevin Cheang, Geoff Hulette, Rahul Kumar et al. (+5 more)

Language models (LMs) are increasingly used to interact with external services via programs written in domain-specific languages (DSLs). Unfortunately, since DSLs are often low-resource and esoteric, LMs frequently produce syntactically invalid programs in these languages. Grammar-constrained decoding can eliminate such failures, but requires syntactic constraints. These are usually in the form of a context-free grammar for the target language, an artifact that is hard to come by for third-party DSLs. In this work, we define an agent, called Autogrammar, that automatically learns context-free grammars from documentation and execution data. Autogrammar is formalized as a Kripke structure whose nondeterministic choices are resolved by a language model, enabling declarative control of agent behavior via linear temporal logic constraints. We evaluate four versions of Autogrammar on three DSLs (i.e., Amazon CloudWatch Logs Insights, Dynatrace Query Language, and Datadog Search Syntax) and find that it generates grammars that achieve near perfect precision on unseen data; that temporal restrictions reduce execution time by 3.8x without incurring statistically-significant loss in precision; that execution data is crucial while documentation is dispensable; and that grammar-constrained decoding using Autogrammar-generated grammars significantly improves end-to-end LM performance on eight out of ten real tasks, matching or exceeding the performance of a professionally-maintained grammar. In comparison, the context-free grammars generated by existing LM baselines and a state-of-the-art formal technique perform significantly worse over the same evaluation.
Read the full abstract

All authors Kevin Cheang, Geoff Hulette, Rahul Kumar, Felipe R. Monteiro, Federico Mora, Robin Salkeld, Lin Tan, Serdar Tasiran

Wednesday 5 August

Computation and Language · Cross-listed from cs.IR

RESEARCHHUB: A Personalized Literature Discovery System for Biomedical Preprints

Viswonathan Manoranjan, Taeyoon Kim, Nicholas Do et al. (+5 more)

We present ResearchHub, an interactive system for personalized literature discovery over biomedical preprints. Given a researcher profile or a free-form query, ResearchHub retrieves relevant papers from a domain-specific MedRxiv corpus and helps users move beyond near-duplicate recommendations through controllable novelty, diversity, and curriculum ordering. At the core of the system is prism, a bi-encoder retriever adapted to biomedical preprints using GPT-4o-generated natural language queries and a hybrid Multiple Negatives Ranking plus MarginMSE objective with cross-encoder soft labels. On 44,937 MedRxiv papers from 2023–2025, prism outperforms strong off-the-shelf dense retrievers, improving Recall@10 over SPECTER2-Search by 12.4 percentage points. ResearchHub further supports frontier retrieval based on a user's reading history, DPP-based diversity re-ranking, Local Intrinsic Dimensionality ordering from mainstream to specialized papers, and lightweight attribution cards for result explanation. The full pipeline runs with low latency, making the system suitable
Read the full abstract

All authors Viswonathan Manoranjan, Taeyoon Kim, Nicholas Do, Sang Woo Kim, Dawei Li, Zhen Tan, Tianlong Chen, Yunmei Liu

Computation and Language

Rewarding Efficient Reasoning Improves Abstention on Underspecified Tasks in Reasoning Models

Polina Tsvilodub, Max Höth, Michael Franke et al. (+2 more)

While modern large reasoning models (LRMs) excel at providing correct answers in many tasks, we provide additional evidence that they often struggle with a critical capability: knowing when to abstain from answering. We analyze this gap by comparing LRM behavior to results from an exploratory human study, revealing that human reasoning effort on unanswerable tasks isn't different from answerable tasks, whereas LRMs waste computational resources by generating longer Chains of Thought (CoTs) on unanswerable than on answerable prompts. To overcome this inefficiency, we take inspiration from a resource-rational perspective on human cognition and introduce a novel GRPO reward that encourages efficient reasoning about whether the task contains all the information needed to solve it. Fine-tuning several 4B and an 8B LRM with this reward leads to human-like abstention performance gains (+14.7% on average) while retaining answering capabilities and boosting the models' efficiency (44% shorter CoTs on average).
Read the full abstract

All authors Polina Tsvilodub, Max Höth, Michael Franke, Björn Deiseroth, Carina Kauf

Computation and Language

Reading Less While Writing: A Closed-Form Bandwidth Dial for Streaming Multimodal Decoders

Yasir Mehmood and Kashif Javed

A decoder that turns video or audio into text conventionally consumes the entire input before emitting a word. Offline this is merely more than the task requires; live it is impossible, since a caption cannot wait for a match to end. Streaming systems bolt on a fixed rule such as wait-$k$, which waits for the same number of input tokens before every word, regardless of the input's length or pace. We replace the fixed offset with ZENDAYA, a schedule governed by a single continuous parameter $\gamma$. It makes the visible source prefix a closed-form function of generation progress, scaled by the input's own predicted length, so an ordinary offline decoder and a real-time streaming decoder become two endpoints of one family rather than separate models. The same scalar fixes, in closed form, the mean fraction of source consumed per emitted word, $\bar{E}(\gamma) \approx 1/(1+\gamma)$, making it at once a latency dial and an interpretable budget. We prove a structural dependency theorem: under any schedule fixed in advance and non-decreasing, no emitted token can depend on input that has not yet arrived. The guarantee holds for trained and untrained weights alike, and extends to unbounded streams under arbitrary asynchronous arrival. The empirical result is counterintuitive: seeing less can produce better text, because a flood of source dilutes attention exactly when the model has the least of its own output to anchor on. Trained from scratch across two modalities and three public corpora (Charades-STA, ActivityNet Captions, LibriHeavy), a compact 29M-parameter decoder matches or beats the fixed schedule while reading less of the source, with the sharpest gains at the lowest latencies, where a fixed offset collapses. Streaming METEOR gains are statistically significant on all three corpora.
Read the full abstract

Computation and Language

Boosting Deepresearch and LongContext Ability with Self-Generated Deepresearch Rollouts Traces

Zihan Wang, Hao Wang, Boyuan Jiang et al. (+8 more)

Deepresearch (DR) agents interact with real-world web environments through multi-turn search and visit, causing their contexts to grow rapidly over time. We observe that, even after DR Agentic Reinforcement Learning (DR-RL), 61.6% of the model's remaining prediction errors can still be attributed to insufficient long-context understanding, including longcontext hallucination and failures in cross-document evidence integration. It motivates us to further break the bottleneck of DR-RL by strengthening the model's long-context ability. However, effective LongContext training requires more than simply increasing context length. To bridge the data gap, we propose `DR Rollouts to LongContext-QA (DR-to-Long)'. The method repurposes DR-RL trajectories, which naturally contain search histories, visited webpages, evidence snippets, and final-answer supervision. It then replaces the compact snippets and webpage summaries in each trajectory with the full contents of their corresponding URLs, producing substantially longer multi-document contexts while preserving the original evidence relationships. Building on DR-to-Long, we introduce DLD (DR -> LongQA -> DR)-RL. DLD-RL first performs a short DR-RL stage to collect rollout trajectories, which are then converted into LongQA instances at zero annotation cost. The model is subsequently optimized with LongQA-RL to strengthen LongContext ability, followed by full DR-RL to continue improving its DR capability. Experiments show that DLD-RL outperforms standard DR-RL by 7.3% on three Deepresearch benchmarks and improves performance by 13.5% on three long-context benchmarks.
Read the full abstract

All authors Zihan Wang, Hao Wang, Boyuan Jiang, Yiqun Zhang, Shi Feng, Xiaocui Yang, Yiwen Ye, Jianghang Lin, Xiaozhong Ji, Jinghao Lin, Kai Wu

Computation and Language · Cross-listed from cs.SI

Whose Judgments Count? Representation Gaps in Crowdsourced Content Moderation Produce Unequal Protection from Perceived Toxicity

Zhaodi Chen and Byungkyu Lee

Content moderation is a central form of digital governance, yet people disagree over what content should be removed from shared online spaces. While platforms aggregate human judgments to build moderation systems, it remains unclear how this process shapes which users are protected from content they perceive as toxic. We address this gap by combining large-scale judgment data with counterfactual simulations that trace how the demographic composition of moderator pools shapes the distribution of protection across users. Applying this framework to removal judgments from 16,221 U.S. respondents evaluating 102,463 comments from Twitter, Reddit, and 4chan, we find demographic heterogeneities in moderation demand. We further reveal a consistent pattern of in-group protection: reductions in perceived toxicity accrue disproportionately to users who share the demographic identities of the moderator pool. Crucially, moderator pools that mirror the demographic composition of self-identified moderators on Prolific widen these disparities relative to a nationally representative baseline, while even fully representative pools fail to ensure equal protection: Black and LGB users remain underprotected unless they are represented well beyond their population share. These findings show that unequal protection from perceived toxicity can arise structurally from the aggregation of stratified removal standards, making the demographic composition of moderation inputs a key determinant of who is protected online.
Read the full abstract

Computation and Language · Cross-listed from cs.AI

AutoScientist-Quant: Self-Evolving Coding Agents for Automatic Research in Quantitative Investment

Zongqian Li, Yaoyiran Li, Yaohui Guo et al. (+3 more)

Large language model agents can discover alphas, yet current methods have three weaknesses. The search cannot adapt during the run, automation usually ends at alpha generation while library selection and model choice stay manual, and alpha discovery can read the test window through loop feedback or code problems. We present AutoScientist-Quant, a self evolving search process that regards quantitative research as one budgeted search problem. A single controller conditions every decision on the remaining budget, choosing at each round whether to improve, combine, pivot, or stop, which node to expand, how many alphas to generate, and how to retrieve past trajectories from the shared memory. The same core then selects from the library and tunes the model, closing the loop from hypothesis to deployable strategy. We also review the evaluation pipeline reused from prior work, fix two lookahead problems, and keep the feedback window disjoint from the held out test window, so every comparison tests true generalization. On CSI universes, the framework attains the best value of nearly every metric in every setting, and these conclusions hold across several backbones and markets.
Read the full abstract

All authors Zongqian Li, Yaoyiran Li, Yaohui Guo, Ming Zhang, Nigel Collier, Eugene Ie

Computation and Language

Automatic or Controlled? Repetition Priming Reveals Divergent Processing in Base LLMs, Instruct LLMs, and Humans

Jinglei Ren and Yuyue Wang

Words recur constantly in natural language use, yet it remains unclear whether language models reactivate prior representations or re-evaluate repeated words afresh, and whether post-training changes this default behavior. We apply repetition priming (Shiffrin and Schneider, 1977) to 15 models across five model families (1.5B-14B parameters) in two tasks, semantic categorization and cloze completion, with matched human experiments using identical stimuli. We find that base models exhibit automatic processing: they show immediate facilitation that remains stable across lags, partially survives context removal, and correlates with attention to prior occurrences. Instruct models exhibit controlled processing: their facilitation decays with lag, collapses without expected context, and reverses to interference at larger scales. Within the Qwen 2.5 family, this dissociation increases monotonically with model scale, suggesting that post-training progressively alters repetition processing. Humans show a hybrid profile, with lag-sensitive facilitation resembling instruct models but without interference, suggesting that neither model type fully captures human cognition. Our findings reveal a qualitative shift in how language models process repeated information after post-training and provide mechanistic evidence for the divergence between model behaviors.
Read the full abstract

Computation and Language

Recovering Lesion Parameters from Aphasic Picture Naming Error Profiles in Large Language Models

Yong Yang, Roger Newman-Norlund, Xiang Guan et al. (+10 more)

Interpretability methods for large language models (LLMs) describe internal state but do not directly test whether that state is causally sufficient to produce the observed behavior. In earlier work, we lesioned LLMs to produce error profiles in picture naming, a central task for assessing aphasia, and found that specific lesions produced errors resembling those of individual stroke survivors. Here we ask the inverse question: given an error profile, can the lesion parameters that produced it be recovered, and what does this inverse problem reveal about transformer computation? Lesions in LLaVA-Vicuna 13B were parameterized by layer index, modification percentage, and noise sigma across 4,840 configurations, and error profiles were characterized by a seven-category clinical taxonomy (correct, semantic, unrelated, formal, mixed, neologism, no-response). We trained a multi-task neural network to map error profiles back to perturbation parameters. The problem admitted a partial solution: across 10 independently trained inverse models, modification percentage and noise sigma were recoverable, whereas layer index was recoverable only within a neighborhood. In counterfactual validation, a fresh model instance perturbed with the recovered parameters reproduced the target behavior in 81.4% of cases. This dissociation between low layer recovery and high counterfactual fidelity is consistent with functional redundancy across transformer layers, a property not captured by standard interpretability methods. As an out-of-distribution test, we applied the trained model to picture-naming error profiles from 278 stroke survivors; recovered parameters were syndrome-discriminative, most strongly for perturbation intensity, indicating generalization beyond the training distribution. Counterfactual validation provides a general framework for LLM interpretability claims beyond inverse mapping.
Read the full abstract

All authors Yong Yang, Roger Newman-Norlund, Xiang Guan, Saeed Ahmadi, Regan Willis, Nadra Salman, Kalil Warren, Sophie Arheix-Parras, Srihari Nelakuditi, Leonardo Bonilha, Christopher Rorden, Rutvik H. Desai, Julius Fridriksson

Computation and Language

NTDH: Complex Reasoning for Comprehensive Affective Analysis

Tianlei Zhu, Zhiwei Liu, Yuyan Wang et al. (+2 more)

Comprehensive affective analysis is challenging for two reasons: it spans heterogeneous prediction tasks with continuous, ordinal, and multi-label outputs, and affective meaning is context-dependent, requiring conflicting cues to be reconciled rather than mapped directly to labels. Existing methods learn this mapping directly and do not model the reconciliation explicitly. We recast the task as a complex-reasoning problem, which yields one output interface across heterogeneous label spaces and a trajectory over which a verifiable reward can be optimised; to our knowledge, this is the first such treatment covering both sentiment and emotion. The obstacle is on the data side: affective reasoning traces must be synthesised, and generic synthesis is misaligned with the targets, tolerances, and phenomena of affect, and discards or leaks its failure cases. We propose NTDH, which addresses these four failures. Naturalisation sets the training answer to the gold label, so it is correct by construction. A Tolerance-aware gate checks each answer against the task's own scoring margin. Domain-aware strategies refine the reasoning using ideas from affective science. Directional Hints report only the type and direction of an error, without exposing the target. We train Qwen3-8B with SFT and then GRPO under the same tolerance used for verification (up to a more permissive construction gate on the multi-label subtask), and a component ablation quantifies the data-quality effect of each part. Using 16,302 training records, about 14x fewer than comparable instruction-tuned systems, the final policy improves over its SFT checkpoint on five of six official-test metrics and achieves the strongest EI-reg result among the compared systems, at a Pearson correlation of 0.862.
Read the full abstract

All authors Tianlei Zhu, Zhiwei Liu, Yuyan Wang, Xiao-Yang Liu, Sophia Ananiadou

Computation and Language · Cross-listed from cs.SD

Multi Codec Discrete Diffusion Model for Text Guided Speech Inpainting and Editing

Iftach Shoham, Tali Dror, Oren Gal et al. (+3 more)

Speech recordings often contain missing, corrupted, or incorrect regions that must be reconstructed or modified without re-synthesizing the entire utterance. Speech inpainting restores missing segments, whereas speech editing replaces spoken content according to an edited transcript. Both tasks require the generated speech to express the intended words while remaining consistent with the surrounding speaker identity, prosody, timing, and recording conditions. Discrete diffusion is particularly well suited to these tasks because it can iteratively refine masked tokens while jointly conditioning on both left and right acoustic context. We introduce SIEDD, a discrete diffusion framework for text-guided speech inpainting and editing over hierarchical codec tokens. Its core architecture, HiCoDD, follows the RVQ generation order by representing previously generated codebooks as clean, committed acoustic context and applying diffusion only to the current refinement codebook. This separation enables leakage-free joint training while matching sequential coarse-to-fine inference. The model further combines phoneme-level conditioning, span-localized classifier-free guidance, and duration prediction to support both fixed-duration inpainting and variable-duration text edits. On the RealEdit benchmark, SIEDD achieves the best overall speech-editing performance among the evaluated methods. It also outperforms the evaluated autoregressive baselines across all speech-inpainting settings, on both single and multiple gaps. These results demonstrate that explicitly modeling the codec hierarchy substantially improves context-preserving speech reconstruction and editing. See our full code at https://github.com/iftachShoham/SIEDD.
Read the full abstract

All authors Iftach Shoham, Tali Dror, Oren Gal, Haim Permuter, Gilad Katz, Eliya Nachmani

Computation and Language · Cross-listed from cs.LG

Latent Fact-Checking: Detecting Misinformation through Activation Engineering

Pedro T. Barcelos, Otávio Parraga, Marcelo M. Delucis et al. (+3 more)

The proliferation of misinformation online has driven demand for scalable detection systems. While most existing approaches rely on surface-level linguistic features or external knowledge retrieval, we examine truthfulness as a geometric property of a language model's representation space. We introduce a misinformation detection framework grounded in activation engineering, which leverages the latent geometry of transformer models. Our approach elicits a misinformation direction in the residual stream by contrasting activations from paired truthful and false statements, following the difference-in-means principle of Contrastive Activation Addition (CAA). At inference time, the last-token activation of an unseen claim is projected onto this direction, and the projected representation is fed to an Multilayer Perceptron (MLP) for classification. The procedure requires no fine-tuning of the backbone model, no external evidence retrieval, and no task-specific supervision beyond the contrastive pairs used to estimate the direction. We evaluate the method across 11 models from the Gemma, Llama, and Qwen families, ranging from 270M to 12B parameters, on three fact-checking benchmarks: AVeriTeC, LIAR, and FACTors. The falsehood direction is recoverable across model scales and architectural families, and last-token projection matches or surpasses zero-shot and few-shot prompting baselines on LIAR and FACTors, with the largest gains observed for smaller models. Performance on AVeriTeC is more limited, which we attribute to its evidence-grounded labeling scheme. These findings provide evidence that truthfulness is a structured, linearly separable concept in the latent space of pretrained language models, and point toward interpretability-driven misinformation detection as a practical complement to retrieval-based pipelines. The code is available on https://github.com/Malta-Lab/LaFaCt.
Read the full abstract

All authors Pedro T. Barcelos, Otávio Parraga, Marcelo M. Delucis, Lucas M. Fraga, Lucas S. Kupssinskü, Rodrigo C. Barros

Computation and Language · Cross-listed from cs.GR

GenGA: Editable and Data-Grounded Graphical Abstract Generation for Academic Papers

Takuro Kawada, Shunsuke Kitada and Hitoshi Iyatomi

Graphical Abstracts (GAs) visually summarize the key findings of academic papers, playing a crucial role in facilitating the understanding of research content. Recently, advancements in vision-language models and image generation models have enabled the automatic generation of scientific figures based on paper content. However, most conventional methods output the generated results as raster graphics, making post-editing (e.g., text modification and layout changes) highly difficult. This poses a significant challenge, as they are unsuitable for the iterative figure revision process inherent in paper writing and peer review. To tackle these challenges, we define the novel task of generating editable GAs from paper content and propose GenGA, a new GA generation framework that directly produces figures in vector format. By generating figures as a collection of vector elements with a hierarchical structure, GenGA produces outputs that can be seamlessly imported into existing drawing tools for intuitive, element-level editing. Furthermore, we introduce the Structural Independence Coefficient (SIC), a metric that quantifies the editing simplicity of a figure based on the degree to which local modifications propagate to other elements. Experimental results show that GenGA achieves superior editing simplicity compared to conventional methods, and even surpasses human-authored GAs in conciseness and semantic alignment. We also validate SIC as an effective metric correlated with manual editing costs. This study fundamentally redefines GA generation as an editable vector graphic generation problem grounded in the practical workflows of researchers, significantly promoting effective scientific communication.
Read the full abstract

Computation and Language

DBLAST: Dependent Block Drafting for Stochastic Speculative Decoding

Amirmohammad Karimi, Chao Gao and Negar Hassanpour

Speculative decoding accelerates large language models' inference by using a lightweight drafter to propose multiple future tokens and a target model to verify them. While recent block and diffusion-style drafters can predict several positions in a single pass, their training and sampling procedures are typically optimized for greedy decoding or assume that positions in the draft block are conditionally independent. This assumption becomes brittle in non-greedy speculative decoding, where the target distribution is deliberately stochastic and multiple continuations become plausible. We study this mismatch for block diffusion drafters and show that the accepted draft length degrades as the entropy of the target sampling distribution increases. We propose a dependent block drafter based on a low-rank latent mixture over token positions, complemented by an acceptance-oriented training objective that directly targets the expected verified length. Experiments with Qwen3-4B and Qwen3-8B on GSM8K, MT-Bench, HumanEval, and creative-writing benchmarks show that our approach, namely DBLast, consistently improves accepted length over independent block sampling, especially in higher-entropy decoding regimes.
Read the full abstract

Computation and Language

Example-Guided Prompting for Document-Level Text Simplification

Marina Litvak, Ariel Perstin, Ilan Shtilman et al. (+1 more)

Document-level text simplification requires large language models (LLMs) to rewrite complex documents while preserving meaning, readability, and discourse coherence. Although prompt-based LLMs have shown promising performance, they often produce inconsistent simplifications because textual instructions alone provide limited guidance for complex document-level transformations. We investigate whether retrieved document-simplification examples can improve document-level generation by augmenting prompts with examples selected from a parallel simplification corpus. This example-guided prompting approach enables LLMs to exploit relevant simplification patterns without task-specific fine-tuning. Experiments on the OneStopEnglish corpus using multiple state-of-the-art LLMs show that incorporating retrieved examples consistently improves simplification quality over prompt-only generation and achieves competitive or superior performance compared with representative supervised (T5) and planning-based (PlanSimp) document simplification systems. Furthermore, we find that the benefits of example-guided prompting vary across LLMs, suggesting that effective use of retrieved examples depends on a model's ability to integrate contextual information during generation.
Read the full abstract

All authors Marina Litvak, Ariel Perstin, Ilan Shtilman, Michael Färber

Computation and Language · Cross-listed from cs.LG

EvoHarness-RL: Learning Runtime Harness Coordination for Self-Evolving Agents

Xuying Ning, Dongqi Fu, Tianxin Wei et al. (+17 more)

Long-horizon LLM agents increasingly rely on external execution support to maintain state, track progress, recover from failures, and reuse experience across extended interactions. Yet existing harnesses and their use are often tailored to environments and controlled through prompts, heuristics, or system-specific rules, making agent and harness coordination difficult to jointly optimize. We introduce EvoHarness-RL, a unified framework that separates environment-specific harness implementations from a shared policy-facing interface. EvoHarness-RL organizes external support into a Belief, Progress, and Experience (BPE) workspace and exposes four compact harness actions for accessing and updating this state. We first instantiate BPE as an inference-time scaffold and then make harness coordination learnable through supervised initialization followed by cost-aware GRPO. Across heterogeneous long-horizon tasks, EvoHarness-Base improves the average success rate of frontier models by 10.0 percentage points, while EvoHarness-RL outperforms the strongest open-source baseline by 8.5 percentage points, together with higher RL rollout efficiency and stronger generalization to unseen tasks. Our analyses show that training gradually shifts agents from frequent scaffold use toward selective, environment-dependent harness access as the policy becomes more capable, while the external workspace continues to evolve and refine itself to better support task execution and generalization. Together, these results show that long-horizon agents benefit not only from external scaffolding itself, but also from learning how and when to coordinate with external support as part of a cost-aware policy.
Read the full abstract

All authors Xuying Ning, Dongqi Fu, Tianxin Wei, Yuanchen Bei, Xiyuan Yang, Wujiang Xu, Yueqi Song, Bingxuan Li, Zihao Li, Hanqing Zeng, Xiang Shen, Yajuan Wang, Yifan Wu, Qifan Wang, Jiayi Liu, Hong Li, Yinglong Xia, Xiangjun Fan, Hanghang Tong, Jingrui He

Computation and Language

Mood Matters: How Syntactic Sensitivity Undermines Safety Alignment

Alina Klerings, Jannik Brinkmann, Heiner Stuckenschmidt et al. (+1 more)

Large language models typically undergo post-training to align them with safety policies but there exist many sophisticated jailbreaks that sidestep established safeguards. For instance, prior work by Andriushchenko et al. (2025) has found that changing the grammatical tense from present to past can be enough to elicit harmful responses. In this work, we uncover a more general failure of non-imperative syntactic forms. We demonstrate that this syntactic vulnerability exists in 16 models up to 70B parameters, using behavioral evaluation. To investigate the root cause, we apply causal mediation analysis, finding that refusal is partially conditioned on upstream syntactic features. By steering these purely syntactic features we are able to trigger and suppress refusal. Finally, we trace this ill-conditioning to linguistically biased post-training data of open-source models and show that increasing syntactic diversity can mitigate the issue. Our findings suggest that current alignment approaches introduce confounders that prevent a pure semantic grounding of the refusal decision.
Read the full abstract

All authors Alina Klerings, Jannik Brinkmann, Heiner Stuckenschmidt, Simone Paolo Ponzetto