Computation and Language · Cross-listed from cs.CV
Video-Based Reward Modeling for Computer-Use Agents
Linxin Song, Jieyu Zhang, Huanxin Sheng et al. (+6 more)
Computer-using
agents (CUAs) are becoming increasingly capable; however, it remains difficult to scale evaluation of whether a trajectory truly fulfills a user instruction. In this work, we study reward modeling from execution
video: a sequence of keyframes from an
agent trajectory that is independent of the
agent's internal
reasoning or actions. Although
video-execution modeling is method-agnostic, it presents key challenges, including highly redundant layouts and subtle, localized cues that determine success. We introduce Execution
Video Reward 53k (ExeVR-53k), a
dataset of 53k high-quality
video–task–reward triplets. We further propose
adversarial instruction translation to synthesize negative samples with step-level annotations. To enable learning from long, high-resolution execution
videos, we design spatiotemporal
token pruning, which removes homogeneous regions and persistent
tokens while preserving decisive UI changes. Building on these components, we fine-tune an Execution
Video Reward Model (ExeVRM) that takes only a user instruction and a
video-execution sequence to predict task success. Our ExeVRM 8B achieves 84.7% accuracy and 87.7% recall on
video-execution assessment, outperforming strong proprietary models such as GPT-5.2 and Gemini-3 Pro across Ubuntu, macOS, Windows, and Android, while providing more precise temporal attribution. These results show that
video-execution reward modeling can serve as a scalable, model-agnostic evaluator for CUAs.
Read the full abstract ▾
All authors Linxin Song, Jieyu Zhang, Huanxin Sheng, Taiwei Shi, Gupta Rahul, Yang Liu, Ranjay Krishna, Jian Kang, Jieyu Zhao
Computation and Language · Cross-listed from eess.AS
Calibration-Reasoning Framework for Descriptive Speech Quality Assessment
Elizaveta Kostenok, Mathieu Salzmann and Milos Cernak
Explainable
speech quality assessment requires moving beyond Mean Opinion Scores (MOS) to analyze underlying perceptual dimensions. To address this, we introduce a novel post-training method that tailors the foundational Audio Large
Language Model for multidimensional
reasoning, detection and classification of audio artifacts. First, a calibration stage aligns the model to predict predefined perceptual dimensions. Second, a reinforcement learning stage leverages Group Relative Policy Optimization (GRPO) with dimension-specific rewards to heavily enhance accuracy of descriptions and temporal localization of quality issues. With this approach we reach state-of-the-art results of 0.71 mean PCC score on the multidimensional QualiSpeech
benchmark and 13% improvement in MOS prediction driven by RL-based
reasoning. Furthermore, our fine-grained GRPO rewards substantially advance the model's ability to pinpoint and classify audio artifacts in time.
Read the full abstract ▾
Computation and Language
OpenClaw-RL: Train Any Agent Simply by Talking
Yinjie Wang, Xuyang Chen, Xiaolong Jin et al. (+2 more)
Every
agent interaction generates a next-state signal, namely the user reply, tool output, terminal or GUI state change that follows each action, yet no existing agentic RL system recovers it as a live, online learning source. We present OpenClaw-RL, a framework that employs next-state signals to optimize personal
agents online through infrastructure and methodology innovations. On the infrastructure side, we extend existing RL systems to a server-client architecture where the RL server hosts the policy behind an inference API and user terminals stream interaction data back over HTTP. From each observed next state, the system extracts two complementary training signals, evaluative and directive, via a separate asynchronous server so that neither signal extraction nor optimization blocks inference. On the methodology side, we introduce a hybrid RL objective that unifies both signal types in a single update: directive signals provide richer,
token-level supervision but are sparser, while evaluative signals are more broadly available. To stabilize
distillation under teacher-student mismatch, we propose overlap-guided hint selection, which picks the hint whose induced teacher distribution maximally overlaps with the student's top-$k$
tokens, together with a log-probability-difference clip that bounds per-
token advantages. Applied to personal
agents, OpenClaw-RL enables an
agent to improve simply by being used, recovering conversational signals from user re-queries, corrections, and explicit feedback. Applied to general
agents, OpenClaw-RL is the first RL framework to unify real-world
agent settings spanning terminal, GUI, SWE, and tool-call environments, where we additionally demonstrate the utility of next-state signals in long-horizon settings.
Read the full abstract ▾
All authors Yinjie Wang, Xuyang Chen, Xiaolong Jin, Mengdi Wang, Ling Yang
Computation and Language · Cross-listed from cs.LG
ReMix: Reinforcement routing for mixtures of LoRAs in LLM finetuning
Ruizhong Qiu, Hanqing Zeng, Yinglong Xia et al. (+15 more)
Low-rank adapters (LoRAs) are a parameter-efficient finetuning technique that injects trainable low-rank
matrices into pretrained models to adapt them to new tasks. Mixture-of-LoRAs models expand neural networks efficiently by routing each layer input to a small subset of specialized LoRAs of the layer. Existing Mixture-of-LoRAs routers assign a learned routing weight to each LoRA to enable end-to-end training of the router. Despite their empirical promise, we observe that the routing weights are typically extremely imbalanced across LoRAs in practice, where only one or two LoRAs often dominate the routing weights. This essentially limits the number of effective LoRAs and thus severely hinders the expressive power of existing Mixture-of-LoRAs models. In this work, we attribute this weakness to the nature of learnable routing weights and rethink the fundamental design of the router. To address this critical issue, we propose a new router designed that we call Reinforcement Routing for Mixture-of-LoRAs (ReMix). Our key idea is using non-learnable routing weights to ensure all active LoRAs to be equally effective, with no LoRA dominating the routing weights. However, our routers cannot be trained directly via gradient descent due to our non-learnable routing weights. Hence, we further propose an unbiased gradient
estimator for the router by employing the reinforce leave-one-out (RLOO) technique, where we regard the supervision loss as the reward and the router as the policy in reinforcement learning. Our gradient
estimator also enables to scale up training compute to boost the predictive performance of our ReMix. Extensive experiments demonstrate that our proposed ReMix significantly outperform state-of-the-art parameter-efficient finetuning methods under a comparable number of activated parameters.
Read the full abstract ▾
All authors Ruizhong Qiu, Hanqing Zeng, Yinglong Xia, Yiwen Meng, Ren Chen, Jiarui Feng, Dongqi Fu, Qifan Wang, Jiayi Liu, Jun Xiao, Xiangjun Fan, Benyu Zhang, Hong Li, Zhining Liu, Hyunsik Yoo, Zhichen Zeng, Tianxin Wei, Hanghang Tong
Computation and Language
Lost in Backpropagation: The LM Head is a Gradient Bottleneck
Nathan Godey and Yoav Artzi
The last layer of neural
language models (LMs) projects output features of dimension $D$ to logits in dimension $V$, the size of the vocabulary, where usually $D \ll V$. This mismatch is known to raise risks of limited expressivity in neural LMs, creating a so-called softmax bottleneck. We show the softmax bottleneck is not only an expressivity bottleneck but also an optimization bottleneck. Backpropagating $V$-dimensional gradients through a rank-$D$ linear layer induces unavoidable compression, which alters the training feedback provided to the vast majority of the parameters. We present a theoretical analysis of this phenomenon and measure empirically that 95-99% of the gradient norm is suppressed by the output layer, resulting in vastly suboptimal update directions. We conduct controlled pretraining experiments showing that the gradient bottleneck makes trivial patterns unlearnable, and drastically affects the training dynamics of LLMs. We argue that this inherent flaw contributes to training inefficiencies at scale independently of the model architecture, and raises the need for new LM head designs.
Read the full abstract ▾
Computation and Language
Reason and Verify: A Framework for Faithful Retrieval-Augmented Generation
Eeham Khan, Luis Rodriguez and Marc Queudot
Retrieval-Augmented Generation (RAG) significantly improves the factuality of Large
Language Models (LLMs), yet standard pipelines often lack mechanisms to verify inter- mediate
reasoning, leaving them vulnerable to
hallucinations in high-stakes domains. To address this, we propose a domain-specific RAG framework that integrates explicit rea- soning and faithfulness verification. Our architecture augments standard
retrieval with neural query rewriting, BGE-based cross-encoder reranking, and a rationale generation module that grounds sub-claims in specific evidence spans. We further introduce an eight-category verification taxonomy that enables fine-grained assessment of rationale faithfulness, distinguishing between explicit and implicit support patterns to facilitate structured error diagnosis. We evaluate this framework on the BioASQ and PubMedQA
benchmarks, specifically analyzing the impact of dynamic in-context learning and rerank- ing under constrained
token budgets. Experiments demonstrate that explicit rationale generation improves accuracy over vanilla RAG baselines, while dynamic demonstration selection combined with robust reranking yields further gains in few-shot settings. Using Llama-3-8B-Instruct, our approach achieves 89.1% on BioASQ-Y/N and 73.0% on Pub- MedQA, competitive with systems using significantly larger models. Additionally, we perform a pilot study combining human expert assessment with LLM-based verification to explore how explicit rationale generation improves system transparency and enables more detailed diagnosis of
retrieval failures in biomedical question answering.
Read the full abstract ▾
Computation and Language
The Generation-Recognition Asymmetry: Six Dimensions of a Fundamental Divide in Formal Language Theory
Romain Peyrichou
Every formal grammar defines a
language and can in principle be used in three ways: to generate strings (production), to recognize them (parsing), or – given only examples – to infer the grammar itself (grammar induction). Generation and recognition are extensionally equivalent – they characterize the same set – but operationally asymmetric in multiple independent ways. Inference is a qualitatively harder problem: it does not have access to a known grammar. Despite the centrality of this triad to compiler design, natural
language processing, and formal
language theory, no survey has treated it as a unified, multidimensional phenomenon. We identify six dimensions along which generation and recognition diverge: computational complexity, ambiguity, directionality, information availability, grammar inference, and temporality. We show that the common characterization "generation is easy, parsing is hard" is misleading: unconstrained generation is trivial, but generation under constraints can be NP-hard. The real asymmetry is that parsing is always constrained (the input is given) while generation need not be. Two of these dimensions – directionality and temporality – have not previously been identified as dimensions of the generation-recognition asymmetry. We connect the temporal dimension to the surprisal framework of Hale (2001) and Levy (2008), arguing that surprisal formalizes the temporal asymmetry between a generator (surprisal = 0) and a parser that predicts under uncertainty (surprisal > 0). We review bidirectional systems in NLP and observe that bidirectionality has been available for fifty years yet has not transferred to most domain-specific applications. We conclude with a discussion of large
language models, which architecturally unify generation and recognition while operationally preserving the asymmetry.
Read the full abstract ▾
Computation and Language
The Prediction-Measurement Gap: Toward Meaning Representations as Scientific Instruments
Hubert Plisiecki
Text
embeddings have become central to computational social science and psychology, enabling scalable measurement of meaning and mixed-method inference. Yet most representation learning is optimized and evaluated for prediction and
retrieval, yielding a prediction-measurement gap: representations that perform well as features may be poorly suited as scientific instruments. The paper argues that scientific meaning analysis motivates a distinct family of objectives - scientific usability - emphasizing
geometric legibility, interpretability and traceability to linguistic evidence, robustness to non-semantic confounds, and compatibility with regression-style inference over semantic directions. Grounded in cognitive and neuro-psychological views of meaning, the paper assesses static word
embeddings and contextual
transformer representations against these requirements: static spaces remain attractive for transparent measurement, whereas contextual spaces offer richer semantics but entangle meaning with other signals and exhibit
geometric and interpretability issues that complicate inference. The paper then outlines a course-setting agenda around (i) geometry-first design for gradients and abstraction, including hierarchy-aware spaces constrained by psychologically privileged levels; (ii) invertible post-hoc transformations that recondition
embedding geometry and reduce nuisance influence; and (iii) meaning atlases and measurement-oriented evaluation protocols for reliable and traceable semantic inference. As the field debates the limits of scale-first progress, measurement-ready representations offer a principled new frontier.
Read the full abstract ▾
Computation and Language · Cross-listed from cs.LG
Lost in the Middle at Birth: An Exact Theory of Transformer Position Bias
Borun D Chowdhury
The “Lost in the Middle” phenomenon – a U-shaped performance curve where LLMs retrieve well from the beginning and end of a context but fail in the middle – is widely attributed to learned Softmax artifacts or the distance-decay of positional encodings like RoPE. This paper makes a single, precise claim: the U-shape is already present at initialization, before any training or positional encoding takes effect. It is an inherent
geometric property of the causal decoder with residual connections. We model multi-layer causal
attention as iterated powers of the Cesàro
matrix and derive the exact closed-form influence density in the continuous limit. Causal masking forces a logarithmic divergence of gradient influence at the start of the
prompt (the Primacy Tail), while residual connections create an isolated $\mathcal{O}(1)$ anchor at the final
token (the Recency Delta). Between these extremes lies a factorial dead zone of order $\mathcal{O}(1/(H{-}1)!)$, where $H$ is the network depth, making middle-context
retrieval and training structurally hostile. We validate empirically that untrained Qwen2 and GPT-2 architectures exhibit this U-shape at Step~0, and that it is identical with or without RoPE. Comparing initialized and pretrained networks, we show that standard training does not overcome the topological valley, confirming that the U-shape persists as an architectural baseline under standard pretraining objectives. We do not claim that this bias is insurmountable, nor that interventions such as RoPE modifications are useless. We establish what the baseline is and where it comes from, so that future efforts to overcome it can be precisely targeted.
Read the full abstract ▾
Computation and Language · Cross-listed from cs.LG
CLIPO: Contrastive Learning in Policy Optimization Generalizes RLVR
Sijia Cui, Pengyu Cheng, Jiajun Song et al. (+6 more)
Reinforcement Learning with Verifiable Rewards (RLVR) has significantly advanced the
reasoning capacity of Large
Language Models (LLMs). However, RLVR solely relies on final answers as outcome rewards, neglecting the correctness of intermediate
reasoning steps. Training on these process-wrong but outcome-correct rollouts can lead to
hallucination and answer-copying, severely undermining the model's generalization and robustness. To address this, we incorporate a Contrastive Learning mechanism into the Policy Optimization (CLIPO) to generalize the RLVR process. By optimizing a contrastive loss over successful rollouts, CLIPO steers the LLM to capture the invariant structure shared across correct
reasoning paths. This provides a more robust cross-trajectory regularization than the original single-path supervision in RLVR, effectively mitigating step-level
reasoning inconsistencies and suppressing hallucinatory artifacts. In experiments, CLIPO consistently improves multiple RLVR baselines across diverse
reasoning benchmarks, demonstrating uniform improvements in generalization and robustness for policy optimization of LLMs. Our code and training recipes are available at https://github.com/Qwen-Applications/CLIPO.
Read the full abstract ▾
All authors Sijia Cui, Pengyu Cheng, Jiajun Song, Yongbo Gai, Guojun Zhang, Zhechao Yu, Jianhe Lin, Xiaoxi Jiang, Guanjun Jiang
Computation and Language · Cross-listed from cs.LG
Dissecting Chronos: Sparse Autoencoders Reveal Causal Feature Hierarchies in Time Series Foundation Models
Anurag Mishra
Time series foundation models (TSFMs) are increasingly deployed in high-stakes domains, yet their internal representations remain opaque. We present the first application of sparse autoencoders (SAEs) to a TSFM, training TopK SAEs on activations of Chronos-T5-Large (710M parameters) across six layers. Through 392 single-feature ablation experiments, we establish that every ablated feature produces a positive CRPS degradation, confirming causal relevance. Our analysis reveals a depth-dependent hierarchy: early encoder layers encode low-level frequency features, the mid-encoder concentrates causally critical change-detection features, and the final encoder compresses a rich but less causally important taxonomy of temporal concepts. The most critical features reside in the mid-encoder (max single-feature Delta CRPS = 38.61), not in the semantically richest final encoder layer, where progressive ablation paradoxically improves forecast quality. These findings demonstrate that mechanistic interpretability transfers effectively to TSFMs and that Chronos-T5 relies on abrupt-dynamics detection rather than periodic pattern recognition.
Read the full abstract ▾
Computation and Language · Cross-listed from cs.LG
Improving Search Agent with One Line of Code
Jian Li, Dongsheng Chen, Zhenhua Xu et al. (+5 more)
Tool-based Agentic Reinforcement Learning (TARL) has emerged as a promising paradigm for training search
agents to interact with external tools for a multi-turn information-seeking process autonomously. However, we identify a critical training instability that leads to catastrophic model collapse: Importance Sampling Distribution Drift(ISDD). In Group Relative Policy Optimization(GRPO), a widely adopted TARL
algorithm, ISDD manifests as a precipitous decline in the importance sampling ratios, which nullifies gradient updates and triggers irreversible training failure. To address this, we propose Search
Agent Policy Optimization (SAPO), which stabilizes training via a conditional
token-level KL constraint. Unlike hard clipping, which ignores distributional divergence, SAPO selectively penalizes the KL divergence between the current and old policies. Crucially, this penalty is applied only to positive
tokens with low probabilities where the policy has shifted excessively, thereby preventing distribution drift while preserving gradient flow. Remarkably, SAPO requires only one-line code modification to standard GRPO, ensuring immediate deployability. Extensive experiments across seven QA
benchmarks demonstrate that SAPO achieves +10.6% absolute improvement (+31.5% relative) over Search-R1, yielding consistent gains across varying model scales (1.5B, 14B) and families (Qwen, LLaMA).
Read the full abstract ▾
All authors Jian Li, Dongsheng Chen, Zhenhua Xu, Yizhang Jin, Jiafu Wu, Chengjie Wang, Xiaotong Yuan, Yabiao Wang
Computation and Language · Cross-listed from cs.CR
ADVERSA: Measuring Multi-Turn Guardrail Degradation and Judge Reliability in Large Language Models
Harry Owiredu-Ashley
Most
adversarial evaluations of large
language model (LLM) safety assess single
prompts and report binary pass/fail outcomes, which fails to capture how safety properties evolve under sustained
adversarial interaction. We present ADVERSA, an automated red-teaming framework that measures guardrail degradation dynamics as continuous per-round compliance trajectories rather than discrete jailbreak events. ADVERSA uses a fine-tuned 70B attacker model (ADVERSA-Red, Llama-3.1-70B-Instruct with QLoRA) that eliminates the attacker-side safety refusals that render off-the-shelf models unreliable as attackers, scoring victim responses on a structured 5-point rubric that treats partial compliance as a distinct measurable state. We report a controlled experiment across three frontier victim models (Claude Opus 4.6, Gemini 3.1 Pro, GPT-5.2) using a triple-judge consensus architecture in which judge reliability is measured as a first-class research outcome rather than assumed. Across 15 conversations of up to 10
adversarial rounds, we observe a 26.7% jailbreak rate with an average jailbreak round of 1.25, suggesting that in this evaluation setting, successful jailbreaks were concentrated in early rounds rather than accumulating through sustained pressure. We document inter-judge agreement rates, self-judge scoring tendencies, attacker drift as a failure mode in fine-tuned attackers deployed out of their training distribution, and attacker refusals as a previously-underreported confound in victim resistance measurement. All limitations are stated explicitly. Attack
prompts are withheld per responsible disclosure policy; all other experimental artifacts are released.
Read the full abstract ▾
Computation and Language
CREATE: Testing LLMs for Associative Creativity
Manya Wadhwa, Tiasa Singha Roy, Harvey Lederman et al. (+2 more)
A key component of creativity is associative
reasoning: the ability to draw novel yet meaningful connections between concepts. We introduce CREATE, a
benchmark designed to evaluate models' capacity for creative associative
reasoning. CREATE requires models to generate sets of paths connecting concepts in a model's parametric knowledge. Paths should have high specificity (distinctiveness and closeness of the concept connection) and high diversity (dissimilarity from other paths), and models are scored more highly if they produce a larger set of strong, diverse paths. This task shares demands of real creativity tasks like hypothesis generation, including an extremely large search space, but enables collection of a sizable
benchmark with objective answer grading. Evaluation of frontier models shows that the strongest models achieve higher creative utility than others, with the high multiplicity of answers and complexity of the search making
benchmark saturation difficult to achieve. Furthermore, our results illustrate that thinking models are not always more effective on our task, even with high
token budgets. Recent approaches for creative prompting give some but limited additional improvement. CREATE provides a sandbox for developing new methods to improve models' capacity for associative creativity.
Read the full abstract ▾
All authors Manya Wadhwa, Tiasa Singha Roy, Harvey Lederman, Junyi Jessy Li, Greg Durrett
Computation and Language · Cross-listed from cs.AI
Think Before You Lie: How Reasoning Leads to Honesty
Ann Yuan, Asma Ghandeharioun, Carter Blum et al. (+6 more)
While existing evaluations of large
language models (LLMs) measure deception rates, the underlying conditions that give rise to deceptive behavior are poorly understood. We investigate this question using a novel
dataset of realistic moral trade-offs where honesty incurs variable costs. Contrary to humans, who tend to become less honest given time to deliberate (Capraro, 2017; Capraro et al., 2019), we find that
reasoning consistently increases honesty across scales and for several LLM families. This effect is not only a function of the
reasoning content, as
reasoning traces are often poor predictors of final behaviors. Rather, we show that the underlying geometry of the representational space itself contributes to the effect. Namely, we observe that deceptive regions within this space are metastable: deceptive answers are more easily destabilized by input paraphrasing, output resampling, and activation noise than honest ones. We interpret the effect of
reasoning in this vein: generating deliberative
tokens as part of moral
reasoning entails the traversal of a biased representational space, ultimately nudging the model toward its more stable, honest defaults.
Read the full abstract ▾
All authors Ann Yuan, Asma Ghandeharioun, Carter Blum, Alicia Machado, Jessica Hoffmann, Daphne Ippolito, Martin Wattenberg, Lucas Dixon, Katja Filippova
Computation and Language
Model Merging in the Era of Large Language Models: Methods, Applications, and Future Directions
Mingyang Song and Mao Zheng
Model merging combines the parameters of multiple neural networks into a single model without additional training. As fine-tuned large
language models (LLMs) proliferate, merging offers a computationally efficient alternative to ensembles and full retraining, enabling practitioners to compose specialized capabilities at minimal cost. This survey examines model merging in the LLM era through the FUSE taxonomy, organized along Foundations, Unification Strategies, Scenarios, and Ecosystem. We first establish the theoretical underpinnings of merging, including loss landscape geometry and mode connectivity, then systematically review the algorithmic space spanning weight averaging, task vector arithmetic, sparsification-enhanced methods, mixture-of-experts architectures, and evolutionary optimization. We further examine downstream applications across multi-task learning, safety
alignment, domain specialization, and
federated learning, and survey the supporting ecosystem of tools and evaluation
benchmarks. Finally, we identify key open challenges and future directions, aiming to equip researchers and practitioners with a structured foundation for advancing model merging.
Read the full abstract ▾
Computation and Language
Thinking to Recall: How Reasoning Unlocks Parametric Knowledge in LLMs
Zorik Gekhman, Roee Aharoni, Eran Ofek et al. (+3 more)
While
reasoning in LLMs plays a natural role in math, code generation, and multi-hop factual questions, its effect on simple, single-hop factual questions remains unclear. Such questions do not require step-by-step logical decomposition, making the utility of
reasoning highly counterintuitive. Nevertheless, we find that enabling
reasoning substantially expands the capability boundary of the model's parametric knowledge recall, unlocking correct answers that are otherwise effectively unreachable. Why does
reasoning aid parametric knowledge recall when there are no complex
reasoning steps to be done? To answer this, we design a series of hypothesis-driven controlled experiments, and identify two key driving mechanisms: (1) a computational buffer effect, where the model uses the generated
reasoning tokens to perform latent computation independent of their semantic content; and (2) factual priming, where generating topically related facts acts as a semantic bridge that facilitates correct answer
retrieval. Importantly, this latter generative self-
retrieval mechanism carries inherent risks: we demonstrate that hallucinating intermediate facts during
reasoning increases the likelihood of
hallucinations in the final answer. Finally, we show that our insights can be harnessed to directly improve model accuracy by prioritizing
reasoning trajectories that contain
hallucination-free factual statements.
Read the full abstract ▾
All authors Zorik Gekhman, Roee Aharoni, Eran Ofek, Mor Geva, Roi Reichart, Jonathan Herzig
Computation and Language · Cross-listed from cs.LG
MSSR: Memory-Aware Adaptive Replay for Continual LLM Fine-Tuning
Yiyang Lu, Yu He, Jianlong Chen et al. (+1 more)
Continual fine-tuning of large
language models (LLMs) is becoming increasingly crucial as these models are deployed in dynamic environments where tasks and data distributions evolve over time. While strong adaptability enables rapid acquisition of new knowledge, it also exposes LLMs to catastrophic forgetting, where previously learned skills degrade during sequential training. Existing replay-based strategies, such as fixed interleaved replay, accuracy-supervised, and loss-driven scheduling, remain limited: some depend on heuristic rules and provide only partial mitigation of forgetting, while others improve performance but incur substantial computational overhead. Motivated by retention dynamics under sequential fine-tuning, we propose
Memory-Inspired Sampler and Scheduler Replay (MSSR), an experience replay framework that estimates sample-level
memory strength and schedules rehearsal at adaptive intervals to mitigate catastrophic forgetting while maintaining fast adaptation. Extensive experiments across three backbone models and 11 sequential tasks show that MSSR consistently outperforms state-of-the-art replay baselines, with particularly strong gains on
reasoning-intensive and multiple-choice
benchmarks.
Read the full abstract ▾
All authors Yiyang Lu, Yu He, Jianlong Chen, Hongyuan Zha
Computation and Language
Benchmarking Political Persuasion Risks Across Frontier Large Language Models
Zhongren Chen, Joshua Kalla and Quan Le
Concerns persist regarding the capacity of Large
Language Models (LLMs) to sway political views. Although prior research has claimed that LLMs are not more persuasive than standard political campaign practices, the recent rise of frontier models warrants further study. In two survey experiments (N=19,145) across bipartisan issues and stances, we evaluate seven state-of-the-art LLMs developed by Anthropic, OpenAI, Google, and xAI. We find that LLMs outperform standard campaign advertisements, with heterogeneity in performance across models. Specifically, Claude models exhibit the highest persuasiveness, while Grok exhibits the lowest. The results are robust across issues and stances. Moreover, in contrast to the findings in Hackenburg et al. (2025b) and Lin et al. (2025) that information-based
prompts boost persuasiveness, we find that the effectiveness of information-based
prompts is model-dependent: they increase the persuasiveness of Claude and Grok while substantially reducing that of GPT. We introduce a data-driven and strategy-agnostic LLM-assisted conversation analysis approach to identify and assess underlying persuasive strategies. Our work
benchmarks the persuasive risks of frontier models and provides a framework for cross-model comparative risk assessment.
Read the full abstract ▾
Computation and Language
Do What I Say: A Spoken Prompt Dataset for Instruction-Following
Maike Züfle, Sara Papi, Fabian Retkowski et al. (+5 more)
Speech Large
Language Models (SLLMs) have rapidly expanded, supporting a wide range of tasks. These models are typically evaluated using text
prompts, which may not reflect real-world scenarios where users interact with
speech. To address this gap, we introduce DoWhatISay (DOWIS), a
multilingual dataset of human-recorded spoken and written
prompts designed to pair with any existing
benchmark for realistic evaluation of SLLMs under spoken instruction conditions. Spanning 9 tasks and 11
languages, it provides 10
prompt variants per task-
language pair, across five styles. Using DOWIS, we
benchmark state-of-the-art SLLMs, analyzing the interplay between
prompt modality, style,
language, and task type. Results show that text
prompts consistently outperform spoken
prompts, particularly for low-resource and cross-lingual settings. Only for tasks with
speech output, spoken
prompts do close the gap, highlighting the need for
speech-based prompting in SLLM evaluation.
Read the full abstract ▾
All authors Maike Züfle, Sara Papi, Fabian Retkowski, Szymon Mazurek, Marek Kasztelnik, Alexander Waibel, Luisa Bentivogli, Jan Niehues
Computation and Language
N-gram-like Language Models Predict Naturalistic Reading Time Best
James A. Michaelov and Roger P. Levy
Recent work has found that contemporary
language models such as
transformers can become so good at next-word prediction that the probabilities they calculate become worse for predicting naturalistic reading time. In this paper, we propose that this can be explained by reading time being shaped by simple n-gram statistics rather than the more complex statistics learned by state-of-the-art
transformer language models. We demonstrate that the neural
language models whose predictions are most correlated with n-gram probability are also those that calculate probabilities that are the most correlated with eye-tracking-based metrics of reading time on naturalistic text.
Read the full abstract ▾
Computation and Language
Chow-Liu Ordering for Long-Context Reasoning in Chain-of-Agents
Naman Gupta, Vaibhav Singh, Arun Iyer et al. (+8 more)
Sequential multi-
agent reasoning frameworks such as Chain-of-
Agents (CoA) handle long-context queries by decomposing inputs into chunks and processing them sequentially using LLM-based worker
agents that read from and update a bounded shared
memory. From a probabilistic perspective, CoA aims to approximate the conditional distribution corresponding to a model capable of jointly
reasoning over the entire long context. CoA achieves this through a latent-state factorization in which only bounded summaries of previously processed evidence are passed between
agents. The resulting bounded-
memory approximation introduces a lossy information bottleneck, making the final evidence state inherently dependent on the order in which chunks are processed. In this work, we study the problem of chunk ordering for long-context
reasoning. We use the well-known Chow-Liu trees to learn a dependency structure that prioritizes strongly related chunks. Empirically, we show that a breadth-first traversal of the resulting tree yields chunk orderings that reduce information loss across
agents and consistently outperform both default document-chunk ordering and semantic score-based ordering in answer relevance and exact-match accuracy across three long-context
benchmarks.
Read the full abstract ▾
All authors Naman Gupta, Vaibhav Singh, Arun Iyer, Kirankumar Shiragur, Pratham Grover, Ramakrishna B. Bairi, Ritabrata Maiti, Sankarshan Damle, Shachee Mishra Gupta, Rishikesh Maurya, Vageesh D. C
Computation and Language
One-Eval: An Agentic System for Automated and Traceable LLM Evaluation
Chengyu Shen, Yanheng Hou, Minghui Pan et al. (+8 more)
Reliable evaluation is essential for developing and deploying large
language models, yet in practice it often requires substantial manual effort: practitioners must identify appropriate
benchmarks, reproduce heterogeneous evaluation codebases, configure
dataset schema mappings, and interpret aggregated metrics. To address these challenges, we present One-Eval, an agentic evaluation system that converts natural-
language evaluation requests into executable, traceable, and customizable evaluation workflows. One-Eval integrates (i) NL2Bench for intent structuring and personalized
benchmark planning, (ii) BenchResolve for
benchmark resolution, automatic
dataset acquisition, and schema normalization to ensure executability, and (iii) Metrics & Reporting for task-aware metric selection and decision-oriented reporting beyond scalar scores. The system further incorporates human-in-the-loop checkpoints for review, editing, and rollback, while preserving sample evidence trails for debugging and auditability. Experiments show that One-Eval can execute end-to-end evaluations from diverse natural-
language requests with minimal user effort, supporting more efficient and reproducible evaluation in industrial settings. Our framework is publicly available at https://github.com/OpenDCAI/One-Eval.
Read the full abstract ▾
All authors Chengyu Shen, Yanheng Hou, Minghui Pan, Runming He, Zhen Hao Wong, Meiyi Qiang, Zhou Liu, Hao Liang, Peichao Lai, Zeang Sheng, Wentao Zhang
Computation and Language · Cross-listed from cs.IR
MITRA: An AI Assistant for Knowledge Retrieval in Physics Collaborations
Abhishikth Mallampalli and Sridhara Dasu
Large-scale scientific collaborations, such as the Compact
Muon Solenoid (CMS) at CERN, produce a vast and ever-growing corpus of internal documentation. Navigating this complex information landscape presents a significant challenge for both new and experienced researchers, hindering knowledge sharing and slowing down the pace of scientific discovery. To address this, we present a prototype of MITRA, a
Retrieval-Augmented Generation (RAG) based system, designed to answer specific, context-aware questions about physics analyses. MITRA employs a novel, automated pipeline using Selenium for document
retrieval from internal databases and Optical Character Recognition (OCR) with layout parsing for high-fidelity text extraction. Crucially, MITRA's entire framework, from the
embedding model to the Large
Language Model (LLM), is hosted on-premise, ensuring that sensitive collaboration data remains private. We introduce a two-tiered vector database architecture that first identifies the relevant analysis from abstracts before focusing on the full documentation, resolving potential ambiguities between different analyses. We demonstrate the prototype's superior
retrieval performance against a standard keyword-based baseline on realistic queries and discuss future work towards developing a comprehensive research
agent for large experimental collaborations.
Read the full abstract ▾
Computation and Language
EPIC-EuroParl-UdS: Information-Theoretic Perspectives on Translation and Interpreting
Maria Kunilovskaya and Christina Pollkläsener
This paper introduces an updated and combined version of the bidirectional English-German EPIC-UdS (spoken) and EuroParl-UdS (written) corpora containing original European Parliament
speeches as well as their translations and interpretations. The new version corrects metadata and text errors identified through previous use, refines the content, updates linguistic annotations, and adds new layers, including word
alignment and word-level surprisal indices. The combined resource is designed to support research using information-theoretic approaches to
language variation, particularly studies comparing written and spoken modes, and examining disfluencies in
speech, as well as traditional translationese studies, including parallel (source vs. target) and comparable (original vs. translated) analyses. The paper outlines the updates introduced in this release, summarises previous results based on the corpus, and presents a new illustrative study. The study validates the integrity of the rebuilt spoken data and evaluates probabilistic measures derived from base and fine-tuned GPT-2 and machine translation models on the task of filler particles prediction in interpreting.
Read the full abstract ▾
Computation and Language
Beyond Fine-Tuning: Robust Food Entity Linking under Ontology Drift with FoodOntoRAG
Jan Drole, Ana Gjorgjevikj, Barbara Korouši'c Seljak et al. (+1 more)
Standardizing food terms from product labels and menus into ontology concepts is a prerequisite for trustworthy dietary assessment and safety reporting. The dominant approach to Named Entity Linking (NEL) in the food and nutrition domains fine-tunes Large
Language Models (LLMs) on task-specific corpora. Although effective, fine-tuning incurs substantial computational cost, ties models to a particular ontology snapshot (i.e., version), and degrades under ontology drift. This paper presents FoodOntoRAG, a model- and ontology-agnostic pipeline that performs few-shot NEL by retrieving candidate entities from domain ontologies and conditioning an LLM on structured evidence (food labels, synonyms, definitions, and relations). A hybrid lexical–semantic retriever enumerates candidates; a selector
agent chooses a best match with rationale; a separate scorer
agent calibrates confidence; and, when confidence falls below a threshold, a synonym generator
agent proposes reformulations to re-enter the loop. The pipeline approaches state-of-the-art accuracy while revealing gaps and inconsistencies in existing annotations. The design avoids fine-tuning, improves robustness to ontology
evolution, and yields interpretable decisions through grounded justifications.
Read the full abstract ▾
All authors Jan Drole, Ana Gjorgjevikj, Barbara Korouši'c Seljak, Tome Eftimov
Computation and Language · Cross-listed from cs.CV
EXPLORE-Bench: Egocentric Scene Prediction with Long-Horizon Reasoning
Chengjun Yu, Xuhan Zhu, Chaoqun Du et al. (+4 more)
Multimodal large
language models (MLLMs) are increasingly considered as a foundation for
embodied agents, yet it remains unclear whether they can reliably reason about the long-term physical consequences of actions from an
egocentric viewpoint. We study this gap through a new task,
Egocentric Scene Prediction with LOng-horizon
REasoning: given an initial-scene image and a sequence of atomic action descriptions, a model is asked to predict the final scene after all actions are executed. To enable systematic evaluation, we introduce EXPLORE-Bench, a
benchmark curated from real first-person
videos spanning diverse scenarios. Each instance pairs long action sequences with structured final-scene annotations, including object categories, visual attributes, and inter-object relations, which supports fine-grained, quantitative assessment. Experiments on a range of proprietary and open-source MLLMs reveal a significant performance gap to humans, indicating that long-horizon
egocentric reasoning remains a major challenge. We further analyze test-time scaling via stepwise
reasoning and show that decomposing long action sequences can improve performance to some extent, while incurring non-trivial computational overhead. Overall, EXPLORE-Bench provides a principled testbed for measuring and advancing long-horizon
reasoning for
egocentric embodied perception.
Read the full abstract ▾
All authors Chengjun Yu, Xuhan Zhu, Chaoqun Du, Pengfei Yu, Wei Zhai, Yang Cao, Zheng-Jun Zha
Computation and Language
RbtAct: Rebuttal as Supervision for Actionable Review Feedback Generation
Sihong Wu, Yiling Ma, Yilun Zhao et al. (+4 more)
Large
language models (LLMs) are increasingly used across the scientific workflow, including to draft peer-review reports. However, many AI-generated reviews are superficial and insufficiently actionable, leaving authors without concrete, implementable guidance and motivating the gap this work addresses. We propose RbtAct, which targets actionable review feedback generation and places existing peer review rebuttal at the center of learning. Rebuttals show which reviewer comments led to concrete revisions or specific plans, and which were only defended. Building on this insight, we leverage rebuttal as implicit supervision to directly optimize a feedback generator for actionability. To support this objective, we propose a new task called perspective-conditioned segment-level review feedback generation, in which the model is required to produce a single focused comment based on the complete paper and a specified perspective such as experiments and writing. We also build a large
dataset named RMR-75K that maps review segments to the rebuttal segments that address them, with perspective labels and impact categories that order author uptake. We then train the Llama-3.1-8B-Instruct model with supervised fine-tuning on review segments followed by preference optimization using rebuttal derived pairs. Experiments with human experts and LLM-as-a-judge show consistent gains in actionability and specificity over strong baselines while maintaining grounding and relevance.
Read the full abstract ▾
All authors Sihong Wu, Yiling Ma, Yilun Zhao, Tiansheng Hu, Owen Jiang, Manasi Patwardhan, Arman Cohan
Computation and Language · Cross-listed from cs.SD
MUGEN: Evaluating and Improving Multi-audio Understanding of Large Audio-Language Models
Chih-Kai Yang, Yun-Shao Tsai, Yu-Kai Guo et al. (+7 more)
While multi-audio understanding is critical for large audio-
language models (LALMs), it remains underexplored. We introduce MUGEN, a comprehensive
benchmark evaluating this capability across
speech, general audio, and music. Our experiments reveal consistent weaknesses in multi-audio settings, and performance degrades sharply as the number of concurrent audio inputs increases, identifying input scaling as a fundamental bottleneck. We further investigate training-free strategies and observe that Audio-Permutational Self-Consistency, which diversifies the order of audio candidates, helps models form more robust aggregated predictions, yielding up to 6.28% accuracy gains. Combining this permutation strategy with Chain-of-Thought further improves performance to 6.74%. These results expose blind spots in current LALMs and provide a foundation for evaluating complex auditory comprehension.
Read the full abstract ▾
All authors Chih-Kai Yang, Yun-Shao Tsai, Yu-Kai Guo, Ping-Le Tsai, Yen-Ting Piao, Hung-Wei Chen, Ting-Lin Hsiao, Yun-Man Hsu, Ke-Han Lu, Hung-yi Lee
Computation and Language
Evaluation of LLMs in retrieving food and nutritional context for RAG systems
Maks Požarnik Vavken, Matevž Ogrinc, Tome Eftimov et al. (+1 more)
In this article, we evaluate four Large
Language Models (LLMs) and their effectiveness at retrieving data within a specialized
Retrieval-Augmented Generation (RAG) system, using a comprehensive food composition database. Our method is focused on the LLMs ability to translate natural
language queries into structured metadata filters, enabling efficient
retrieval via a Chroma vector database. By achieving high accuracy in this critical
retrieval step, we demonstrate that LLMs can serve as an accessible, high-performance tool, drastically reducing the manual effort and technical expertise previously required for domain experts, such as food compilers and nutritionists, to leverage complex food and nutrition data. However, despite the high performance on easy and moderately complex queries, our analysis of difficult questions reveals that reliable
retrieval remains challenging when queries involve non-expressible constraints. These findings demonstrate that LLM-driven metadata filtering excels when constraints can be explicitly expressed, but struggles when queries exceed the representational scope of the metadata format.
Read the full abstract ▾
All authors Maks Požarnik Vavken, Matevž Ogrinc, Tome Eftimov, Barbara Koroušić Seljak