# Research Papers — 2025-10 · bundle page 155 of 217

> Full-text markdown of papers #7701–7750 of 10843 published in 2025-10, newest first. AI-readable mirrors of the full 10843-paper 2025-10 archive (50 papers per file).

Canonical page: https://aipod.dev/papers/archive/2025-10/page/155
[← Previous bundle](https://aipod.dev/papers/archive/2025-10/page/154.md)
[Next bundle →](https://aipod.dev/papers/archive/2025-10/page/156.md)

### The Alignment Waltz: Jointly Training Agents to Collaborate for Safety

Authors: Jingyu Zhang, Haozhu Wang, Eric Michael Smith, Sid Wang, Amr Sharaf, Mahesh Pasupuleti, Benjamin Van Durme, Daniel Khashabi, Jason Weston, Hongyuan Zhan
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.08240v2

Harnessing the power of LLMs requires a delicate dance between being helpful and harmless. This creates a fundamental tension between two competing challenges: vulnerability to adversarial attacks that elicit unsafe content, and a tendency for overrefusal on benign but sensitive prompts. Current approaches often navigate this dance with safeguard models that completely reject any content that contains unsafe portions. This approach cuts the music entirely-it may exacerbate overrefusals and fails to provide nuanced guidance for queries it refuses. To teach models a more coordinated choreography, we propose WaltzRL, a novel multi-agent reinforcement learning framework that formulates safety alignment as a collaborative, positive-sum game. WaltzRL jointly trains a conversation agent and a feedback agent, where the latter is incentivized to provide useful suggestions that improve the safety and helpfulness of the conversation agent's responses. At the core of WaltzRL is a Dynamic Improvement Reward (DIR) that evolves over time based on how well the conversation agent incorporates the feedback. At inference time, unsafe or overrefusing responses from the conversation agent are improved rather than discarded. The feedback agent is deployed together with the conversation agent and only engages adaptively when needed, preserving helpfulness and low latency on safe queries. Our experiments, conducted across five diverse datasets, demonstrate that WaltzRL significantly reduces both unsafe responses (e.g., from 39.0% to 4.6% on WildJailbreak) and overrefusals (from 45.3% to 9.9% on OR-Bench) compared to various baselines. By enabling the conversation and feedback agents to co-evolve and adaptively apply feedback, WaltzRL enhances LLM safety without degrading general capabilities, thereby advancing the Pareto front between helpfulness and harmlessness.

---

### Chain-of-Trigger: An Agentic Backdoor that Paradoxically Enhances Agentic Robustness

Authors: Jiyang Qiu, Xinbei Ma, Yunqing Xu, Zhuosheng Zhang, Hai Zhao
Published: 2025-10-09
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.08238v1

The rapid deployment of large language model (LLM)-based agents in real-world applications has raised serious concerns about their trustworthiness. In this work, we reveal the security and robustness vulnerabilities of these agents through backdoor attacks. Distinct from traditional backdoors limited to single-step control, we propose the Chain-of-Trigger Backdoor (CoTri), a multi-step backdoor attack designed for long-horizon agentic control. CoTri relies on an ordered sequence. It starts with an initial trigger, and subsequent ones are drawn from the environment, allowing multi-step manipulation that diverts the agent from its intended task. Experimental results show that CoTri achieves a near-perfect attack success rate (ASR) while maintaining a near-zero false trigger rate (FTR). Due to training data modeling the stochastic nature of the environment, the implantation of CoTri paradoxically enhances the agent's performance on benign tasks and even improves its robustness against environmental distractions. We further validate CoTri on vision-language models (VLMs), confirming its scalability to multimodal agents. Our work highlights that CoTri achieves stable, multi-step control within agents, improving their inherent robustness and task capabilities, which ultimately makes the attack more stealthy and raises potential safty risks.

---

### The Hidden Bias: A Study on Explicit and Implicit Political Stereotypes in Large Language Models

Authors: Konrad Löhr, Shuzhou Yuan, Michael Färber
Published: 2025-10-09
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.08236v2

Large Language Models (LLMs) are increasingly integral to information dissemination and decision-making processes. Given their growing societal influence, understanding potential biases, particularly within the political domain, is crucial to prevent undue influence on public opinion and democratic processes. This work investigates political bias and stereotype propagation across eight prominent LLMs using the two-dimensional Political Compass Test (PCT). Initially, the PCT is employed to assess the inherent political leanings of these models. Subsequently, persona prompting with the PCT is used to explore explicit stereotypes across various social dimensions. In a final step, implicit stereotypes are uncovered by evaluating models with multilingual versions of the PCT. Key findings reveal a consistent left-leaning political alignment across all investigated models. Furthermore, while the nature and extent of stereotypes vary considerably between models, implicit stereotypes elicited through language variation are more pronounced than those identified via explicit persona prompting. Interestingly, for most models, implicit and explicit stereotypes show a notable alignment, suggesting a degree of transparency or "awareness" regarding their inherent biases. This study underscores the complex interplay of political bias and stereotypes in LLMs.

---

### Enhancing Reasoning for Diffusion LLMs via Distribution Matching Policy Optimization

Authors: Yuchen Zhu, Wei Guo, Jaemoo Choi, Petr Molodyk, Bo Yuan, Molei Tao, Yongxin Chen
Published: 2025-10-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.08233v3

Diffusion large language models (dLLMs) are promising alternatives to autoregressive large language models (AR-LLMs), as they potentially allow higher inference throughput. Reinforcement learning (RL) is crucial to enabling dLLMs to achieve performance comparable to that of AR-LLMs on important tasks, such as reasoning. However, RL algorithms well-suited to dLLMs' unique characteristics have yet to be developed. This paper proposes Distribution Matching Policy Optimization (DMPO), a principled and theoretically grounded RL fine-tuning method specifically designed to enhance the reasoning capabilities of dLLMs by matching the dLLM policy distribution to the optimal, reward-tilted one through cross-entropy optimization. We identify a key implementation challenge with small training batch sizes and propose several effective solutions based on a novel weight baseline subtraction technique. DMPO exhibits superior performance on multiple reasoning benchmarks without supervised fine-tuning, achieving up to a $39.63$ percentage-point improvement in accuracy over prior non-DMPO RL baselines and $67.97$ percentage points over the base model, underscoring the effectiveness of the distribution-matching framework. Our code is available at https://github.com/yuchen-zhu-zyc/DMPO.

---

### UAMDP: Uncertainty-Aware Markov Decision Process for Risk-Constrained Reinforcement Learning from Probabilistic Forecasts

Authors: Michal Koren, Or Peretz, Tai Dinh, Philip S. Yu
Published: 2025-10-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.08226v2

Sequential decisions in volatile, high-stakes settings require more than maximizing expected return; they require principled uncertainty management. This paper presents the Uncertainty-Aware Markov Decision Process (UAMDP), a unified framework that couples Bayesian forecasting, posterior-sampling reinforcement learning, and planning under a conditional value-at-risk (CVaR) constraint. In a closed loop, the agent updates its beliefs over latent dynamics, samples plausible futures via Thompson sampling, and optimizes policies subject to preset risk tolerances. We establish regret bounds that converge to the Bayes-optimal benchmark under standard regularity conditions. We evaluate UAMDP in two domains including high-frequency equity trading and retail inventory control, both marked by structural uncertainty and economic volatility. Relative to strong deep learning baselines, UAMDP improves long-horizon forecasting accuracy (RMSE decreases by up to 25% and sMAPE by 32%), and these gains translate into economic performance: the trading Sharpe ratio rises from 1.54 to 1.74 while maximum drawdown is roughly halved. These results show that integrating calibrated probabilistic modeling, exploration aligned with posterior uncertainty, and risk-aware control yields a robust, generalizable approach to safer and more profitable sequential decision-making.

---

### Investigating Counterclaims in Causality Extraction from Text

Authors: Tim Hagen, Niklas Deckers, Felix Wolter, Harrisen Scells, Martin Potthast
Published: 2025-10-09
Categories: cs.CL, cs.LG
arXiv: https://arxiv.org/abs/2510.08224v2

Many causal claims, such as "sugar causes hyperactivity," are disputed or outdated. Yet research on causality extraction from text has almost entirely neglected counterclaims of causation. To close this gap, we conduct a thorough literature review of causality extraction, compile an extensive inventory of linguistic realizations of countercausal claims, and develop rigorous annotation guidelines that explicitly incorporate countercausal language. We also highlight how counterclaims of causation are an integral part of causal reasoning. Based on our guidelines, we construct a new dataset comprising 1028 causal claims, 952 counterclaims, and 1435 uncausal statements, achieving substantial inter-annotator agreement (Cohen's $κ= 0.74$). In our experiments, state-of-the-art models trained solely on causal claims misclassify counterclaims more than 10 times as often as models trained on our dataset.

---

### Selection, Reflection and Self-Refinement: Revisit Reasoning Tasks via a Causal Lens

Authors: Yunlong Deng, Boyang Sun, Yan Li, Lingjing Kong, Zeyu Tang, Kun Zhang, Guangyi Chen
Published: 2025-10-09
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.08222v2

Due to their inherent complexity, reasoning tasks have long been regarded as rigorous benchmarks for assessing the capabilities of machine learning models, especially large language models (LLMs). Although humans can solve these tasks with ease, existing models, even after extensive pre-training and post-training at scale, still fail to perform reasoning reliably. In this paper, we revisit reasoning tasks from a causal perspective, seeking to understand their behavior in latent space and to offer insights for addressing their challenges. Specifically, we cast reasoning tasks as a selection mechanism, in which high-level logical concepts function as selection operators on the given observations, such as, identifying the correct answer in a math problem or filling the appropriate entry in Sudoku. We emphasize two key properties of this formulation that shed light on the difficulty of reasoning tasks. First, the latent space exceeds the observation space in complexity, even when the correct answer is fully determined by the observed input. Second, the latent variables, corresponding to logical thought, are densely structured and exhibit strong dependencies. Building on this formulation, we introduce a framework, called SR$^2$, that incorporates the estimated latent variables as feedback into the selection mechanism, thereby facilitating the learning of dense dependencies among latent representations. The framework consists of three key modules: reflective representation learning, dependency self-refinement, and periodic intermediate alignment. Experimentally, we show that our approach yields significant gains in reasoning accuracy, for example, attaining over 10$\%$ improvement in performance with 8$\times$ fewer parameters on the Sudoku and Maze tasks over the recent advances.

---

### Post-hoc Stochastic Concept Bottleneck Models

Authors: Wiktor Jan Hoffmann, Sonia Laguna, Moritz Vandenhirtz, Emanuele Palumbo, Julia E. Vogt
Published: 2025-10-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.08219v4

Concept Bottleneck Models (CBMs) are interpretable models that predict the target variable through high-level human-understandable concepts, allowing users to intervene on mispredicted concepts to adjust the final output. While recent work has shown that modeling dependencies between concepts can improve CBM performance, especially under interventions, such approaches typically require retraining the entire model, which may be infeasible when access to the original data or compute is limited. In this paper, we introduce Post-hoc Stochastic Concept Bottleneck Models (PSCBMs), a lightweight method that augments any pre-trained CBM with a multivariate normal distribution over concepts by adding only a small covariance-prediction module, without retraining the backbone model. We propose two training strategies and show on real-world data that PSCBMs consistently match or improve both concept and target accuracy over standard CBMs at test time. Furthermore, we show that due to the modeling of concept dependencies, PSCBMs perform much better than CBMs under interventions, while remaining far more efficient than retraining a similar stochastic model from scratch.

---

### Expressive Value Learning for Scalable Offline Reinforcement Learning

Authors: Nicolas Espinosa-Dice, Kiante Brantley, Wen Sun
Published: 2025-10-09
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.08218v1

Reinforcement learning (RL) is a powerful paradigm for learning to make sequences of decisions. However, RL has yet to be fully leveraged in robotics, principally due to its lack of scalability. Offline RL offers a promising avenue by training agents on large, diverse datasets, avoiding the costly real-world interactions of online RL. Scaling offline RL to increasingly complex datasets requires expressive generative models such as diffusion and flow matching. However, existing methods typically depend on either backpropagation through time (BPTT), which is computationally prohibitive, or policy distillation, which introduces compounding errors and limits scalability to larger base policies. In this paper, we consider the question of how to develop a scalable offline RL approach without relying on distillation or backpropagation through time. We introduce Expressive Value Learning for Offline Reinforcement Learning (EVOR): a scalable offline RL approach that integrates both expressive policies and expressive value functions. EVOR learns an optimal, regularized Q-function via flow matching during training. At inference-time, EVOR performs inference-time policy extraction via rejection sampling against the expressive value function, enabling efficient optimization, regularization, and compute-scalable search without retraining. Empirically, we show that EVOR outperforms baselines on a diverse set of offline RL tasks, demonstrating the benefit of integrating expressive value learning into offline RL.

---

### FuelCast: Benchmarking Tabular and Temporal Models for Ship Fuel Consumption

Authors: Justus Viga, Penelope Mueck, Alexander Löser, Torben Weis
Published: 2025-10-09
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.08217v1

In the shipping industry, fuel consumption and emissions are critical factors due to their significant impact on economic efficiency and environmental sustainability. Accurate prediction of ship fuel consumption is essential for further optimization of maritime operations. However, heterogeneous methodologies and limited high-quality datasets hinder direct comparison of modeling approaches. This paper makes three key contributions: (1) we introduce and release a new dataset (https://huggingface.co/datasets/krohnedigital/FuelCast) comprising operational and environmental data from three ships; (2) we define a standardized benchmark covering tabular regression and time-series regression (3) we investigate the application of in-context learning for ship consumption modeling using the TabPFN foundation model - a first in this domain to our knowledge. Our results demonstrate strong performance across all evaluated models, supporting the feasibility of onboard, data-driven fuel prediction. Models incorporating environmental conditions consistently outperform simple polynomial baselines relying solely on vessel speed. TabPFN slightly outperforms other techniques, highlighting the potential of foundation models with in-context learning capabilities for tabular prediction. Furthermore, including temporal context improves accuracy.

---

### SenWave: A Fine-Grained Multi-Language Sentiment Analysis Dataset Sourced from COVID-19 Tweets

Authors: Qiang Yang, Xiuying Chen, Changsheng Ma, Rui Yin, Xin Gao, Xiangliang Zhang
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.08214v1

The global impact of the COVID-19 pandemic has highlighted the need for a comprehensive understanding of public sentiment and reactions. Despite the availability of numerous public datasets on COVID-19, some reaching volumes of up to 100 billion data points, challenges persist regarding the availability of labeled data and the presence of coarse-grained or inappropriate sentiment labels. In this paper, we introduce SenWave, a novel fine-grained multi-language sentiment analysis dataset specifically designed for analyzing COVID-19 tweets, featuring ten sentiment categories across five languages. The dataset comprises 10,000 annotated tweets each in English and Arabic, along with 30,000 translated tweets in Spanish, French, and Italian, derived from English tweets. Additionally, it includes over 105 million unlabeled tweets collected during various COVID-19 waves. To enable accurate fine-grained sentiment classification, we fine-tuned pre-trained transformer-based language models using the labeled tweets. Our study provides an in-depth analysis of the evolving emotional landscape across languages, countries, and topics, revealing significant insights over time. Furthermore, we assess the compatibility of our dataset with ChatGPT, demonstrating its robustness and versatility in various applications. Our dataset and accompanying code are publicly accessible on the repository\footnote{https://github.com/gitdevqiang/SenWave}. We anticipate that this work will foster further exploration into fine-grained sentiment analysis for complex events within the NLP community, promoting more nuanced understanding and research innovations.

---

### LLMs Deceive Unintentionally: Emergent Misalignment in Dishonesty from Misaligned Samples to Biased Human-AI Interactions

Authors: Xuhao Hu, Peng Wang, Xiaoya Lu, Dongrui Liu, Xuanjing Huang, Jing Shao
Published: 2025-10-09
Categories: cs.CL, cs.AI, cs.CR
arXiv: https://arxiv.org/abs/2510.08211v2

Previous research has shown that LLMs finetuned on malicious or incorrect completions within narrow domains (e.g., insecure code or incorrect medical advice) can become broadly misaligned to exhibit harmful behaviors, which is called emergent misalignment. In this work, we investigate whether this phenomenon can extend beyond safety behaviors to a broader spectrum of dishonesty and deception under high-stakes scenarios (e.g., lying under pressure and deceptive behavior). To explore this, we finetune open-sourced LLMs on misaligned completions across diverse domains. Experimental results demonstrate that LLMs show broadly misaligned behavior in dishonesty. Additionally, we further explore this phenomenon in a downstream combined finetuning setting, and find that introducing as little as 1% of misalignment data into a standard downstream task is sufficient to decrease honest behavior over 20%. Furthermore, we consider a more practical human-AI interaction environment where we simulate both benign and biased users to interact with the assistant LLM. Notably, we find that the assistant can be misaligned unintentionally to exacerbate its dishonesty with only 10% biased user population. In summary, we extend the study of emergent misalignment to the domain of dishonesty and deception under high-stakes scenarios, and demonstrate that this risk arises not only through direct finetuning, but also in downstream mixture tasks and practical human-AI interactions. Refer to https://github.com/hxhcreate/LLM_Deceive_Unintentionally for experimental resources.

---

### DODO: Causal Structure Learning with Budgeted Interventions

Authors: Matteo Gregorini, Chiara Boldrini, Lorenzo Valerio
Published: 2025-10-09
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.08207v1

Artificial Intelligence has achieved remarkable advancements in recent years, yet much of its progress relies on identifying increasingly complex correlations. Enabling causality awareness in AI has the potential to enhance its performance by enabling a deeper understanding of the underlying mechanisms of the environment. In this paper, we introduce DODO, an algorithm defining how an Agent can autonomously learn the causal structure of its environment through repeated interventions. We assume a scenario where an Agent interacts with a world governed by a causal Directed Acyclic Graph (DAG), which dictates the system's dynamics but remains hidden from the Agent. The Agent's task is to accurately infer the causal DAG, even in the presence of noise. To achieve this, the Agent performs interventions, leveraging causal inference techniques to analyze the statistical significance of observed changes. Results show better performance for DODO, compared to observational approaches, in all but the most limited resource conditions. DODO is often able to reconstruct with as low as zero errors the structure of the causal graph. In the most challenging configuration, DODO outperforms the best baseline by +0.25 F1 points.

---

### Memory Retrieval and Consolidation in Large Language Models through Function Tokens

Authors: Shaohua Zhang, Yuan Lin, Hang Li
Published: 2025-10-09
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.08203v1

The remarkable success of large language models (LLMs) stems from their ability to consolidate vast amounts of knowledge into the memory during pre-training and to retrieve it from the memory during inference, enabling advanced capabilities such as knowledge memorization, instruction-following and reasoning. However, the mechanisms of memory retrieval and consolidation in LLMs remain poorly understood. In this paper, we propose the function token hypothesis to explain the workings of LLMs: During inference, function tokens activate the most predictive features from context and govern next token prediction (memory retrieval). During pre-training, predicting the next tokens (usually content tokens) that follow function tokens increases the number of learned features of LLMs and updates the model parameters (memory consolidation). Function tokens here roughly correspond to function words in linguistics, including punctuation marks, articles, prepositions, and conjunctions, in contrast to content tokens. We provide extensive experimental evidence supporting this hypothesis. Using bipartite graph analysis, we show that a small number of function tokens activate the majority of features. Case studies further reveal how function tokens activate the most predictive features from context to direct next token prediction. We also find that during pre-training, the training loss is dominated by predicting the next content tokens following function tokens, which forces the function tokens to select the most predictive features from context.

---

### Sentiment Matters: An Analysis of 200 Human-SAV Interactions

Authors: Lirui Guo, Michael G. Burke, Wynita M. Griggs
Published: 2025-10-09
Categories: cs.HC, cs.AI, cs.CL, cs.ET
arXiv: https://arxiv.org/abs/2510.08202v1

Shared Autonomous Vehicles (SAVs) are likely to become an important part of the transportation system, making effective human-SAV interactions an important area of research. This paper introduces a dataset of 200 human-SAV interactions to further this area of study. We present an open-source human-SAV conversational dataset, comprising both textual data (e.g., 2,136 human-SAV exchanges) and empirical data (e.g., post-interaction survey results on a range of psychological factors). The dataset's utility is demonstrated through two benchmark case studies: First, using random forest modeling and chord diagrams, we identify key predictors of SAV acceptance and perceived service quality, highlighting the critical influence of response sentiment polarity (i.e., perceived positivity). Second, we benchmark the performance of an LLM-based sentiment analysis tool against the traditional lexicon-based TextBlob method. Results indicate that even simple zero-shot LLM prompts more closely align with user-reported sentiment, though limitations remain. This study provides novel insights for designing conversational SAV interfaces and establishes a foundation for further exploration into advanced sentiment modeling, adaptive user interactions, and multimodal conversational systems.

---

### The Tournament Tree Method for preference elicitation in Multi-criteria decision-making

Authors: Diego García-Zamora, Álvaro Labella, José Rui Figueira
Published: 2025-10-09
Categories: cs.AI, cs.IT
arXiv: https://arxiv.org/abs/2510.08197v1

Pairwise comparison methods, such as Fuzzy Preference Relations and Saaty's Multiplicative Preference Relations, are widely used to model expert judgments in multi-criteria decision-making. However, their application is limited by the high cognitive load required to complete $m(m-1)/2$ comparisons, the risk of inconsistency, and the computational complexity of deriving consistent value scales. This paper proposes the Tournament Tree Method (TTM), a novel elicitation and evaluation framework that overcomes these limitations. The TTM requires only $m-1$ pairwise comparisons to obtain a complete, reciprocal, and consistent comparison matrix. The method consists of three phases: (i) elicitation of expert judgments using a reduced set of targeted comparisons, (ii) construction of the consistent pairwise comparison matrix, and (iii) derivation of a global value scale from the resulting matrix. The proposed approach ensures consistency by design, minimizes cognitive effort, and reduces the dimensionality of preference modeling from $m(m-1)/2$ to $m$ parameters. Furthermore, it is compatible with the classical Deck of Cards method, and thus it can handle interval and ratio scales. We have also developed a web-based tool that demonstrates its practical applicability in real decision-making scenarios.

---

### Measuring What Matters: The AI Pluralism Index

Authors: Rashid Mushkani
Published: 2025-10-09
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.08193v3

Artificial intelligence systems increasingly mediate knowledge, communication, and decision making. Development and governance remain concentrated within a small set of firms and states, raising concerns that technologies may encode narrow interests and limit public agency. Capability benchmarks for language, vision, and coding are common, yet public, auditable measures of pluralistic governance are rare. We define AI pluralism as the degree to which affected stakeholders can shape objectives, data practices, safeguards, and deployment. We present the AI Pluralism Index (AIPI), a transparent, evidence-based instrument that evaluates producers and system families across four pillars: participatory governance, inclusivity and diversity, transparency, and accountability. AIPI codes verifiable practices from public artifacts and independent evaluations, explicitly handling "Unknown" evidence to report both lower-bound ("evidence") and known-only scores with coverage. We formalize the measurement model; implement a reproducible pipeline that integrates structured web and repository analysis, external assessments, and expert interviews; and assess reliability with inter-rater agreement, coverage reporting, cross-index correlations, and sensitivity analysis. The protocol, codebook, scoring scripts, and evidence graph are maintained openly with versioned releases and a public adjudication process. We report pilot provider results and situate AIPI relative to adjacent transparency, safety, and governance frameworks. The index aims to steer incentives toward pluralistic practice and to equip policymakers, procurers, and the public with comparable evidence.

---

### Training-Free Group Relative Policy Optimization

Authors: Yuzheng Cai, Siqi Cai, Yuchen Shi, Zihan Xu, Lichao Chen, Yulei Qin, Xiaoyu Tan, Gang Li, Zongyi Li, Haojia Lin, Yong Mao, Ke Li et al.
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.08191v1

Recent advances in Large Language Model (LLM) agents have demonstrated their promising general capabilities. However, their performance in specialized real-world domains often degrades due to challenges in effectively integrating external tools and specific prompting strategies. While methods like agentic reinforcement learning have been proposed to address this, they typically rely on costly parameter updates, for example, through a process that uses Supervised Fine-Tuning (SFT) followed by a Reinforcement Learning (RL) phase with Group Relative Policy Optimization (GRPO) to alter the output distribution. However, we argue that LLMs can achieve a similar effect on the output distribution by learning experiential knowledge as a token prior, which is a far more lightweight approach that not only addresses practical data scarcity but also avoids the common issue of overfitting. To this end, we propose Training-Free Group Relative Policy Optimization (Training-Free GRPO), a cost-effective solution that enhances LLM agent performance without any parameter updates. Our method leverages the group relative semantic advantage instead of numerical ones within each group of rollouts, iteratively distilling high-quality experiential knowledge during multi-epoch learning on a minimal ground-truth data. Such knowledge serves as the learned token prior, which is seamlessly integrated during LLM API calls to guide model behavior. Experiments on mathematical reasoning and web searching tasks demonstrate that Training-Free GRPO, when applied to DeepSeek-V3.1-Terminus, significantly improves out-of-domain performance. With just a few dozen training samples, Training-Free GRPO outperforms fine-tuned small LLMs with marginal training data and cost.

---

### R-Horizon: How Far Can Your Large Reasoning Model Really Go in Breadth and Depth?

Authors: Yi Lu, Jianing Wang, Linsen Guo, Wei He, Hongyin Tang, Tao Gui, Xuanjing Huang, Xuezhi Cao, Wei Wang, Xunliang Cai
Published: 2025-10-09
Categories: cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2510.08189v2

Recent trends in test-time scaling for reasoning models (e.g., OpenAI o1, DeepSeek-R1) have led to remarkable improvements through long Chain-of-Thought (CoT). However, existing benchmarks mainly focus on immediate, single-horizon tasks, failing to adequately evaluate models' ability to understand and respond to complex, long-horizon scenarios. To address this incomplete evaluation of Large Reasoning Models (LRMs), we propose R-HORIZON, a method designed to stimulate long-horizon reasoning behaviors in LRMs through query composition. Based on R-HORIZON, we construct a long-horizon reasoning benchmark, comprising complex multi-step reasoning tasks with interdependent problems that span long reasoning horizons. Through comprehensive evaluation of LRMs using the R-HORIZON benchmark, we find that even the most advanced LRMs suffer significant performance degradation. Our analysis reveals that LRMs exhibit limited effective reasoning length and struggle to allocate thinking budget across multiple problems appropriately. Recognizing these limitations, we use R-HORIZON to construct long-horizon reasoning data for reinforcement learning with verified rewards (RLVR). Compared to training with single-horizon data, RLVR with R-HORIZON not only substantially improves performance on the multi-horizon reasoning tasks, but also promotes accuracy on standard reasoning tasks, with an increase of 7.5 on AIME2024. These results position R-HORIZON as a scalable, controllable, and low-cost paradigm for enhancing and evaluating the long-horizon reasoning capabilities of LRMs.

---

### MetricalARGS: A Taxonomy for Studying Metrical Poetry with LLMs

Authors: Chalamalasetti Kranti, Sowmya Vajjala
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.08188v1

Prior NLP work studying poetry has focused primarily on automatic poem generation and summarization. Many languages have well-studied traditions of poetic meter which enforce constraints on a poem in terms of syllable and phoneme patterns. Such advanced literary forms offer opportunities for probing deeper reasoning and language understanding in Large Language Models (LLMs) and their ability to follow strict pre-requisites and rules. In this paper, we introduce MetricalARGS, the first taxonomy of poetry-related NLP tasks designed to evaluate LLMs on metrical poetry across four dimensions: Analysis, Retrieval, Generation, and Support. We discuss how these tasks relate to existing NLP tasks, addressing questions around datasets and evaluation metrics. Taking Telugu as our example language, we illustrate how the taxonomy can be used in practice. MetricalARGS highlights the broader possibilities for understanding the capabilities and limitations of today's LLMs through the lens of metrical poetry.

---

### InstructUDrag: Joint Text Instructions and Object Dragging for Interactive Image Editing

Authors: Haoran Yu, Yi Shi
Published: 2025-10-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.08181v1

Text-to-image diffusion models have shown great potential for image editing, with techniques such as text-based and object-dragging methods emerging as key approaches. However, each of these methods has inherent limitations: text-based methods struggle with precise object positioning, while object dragging methods are confined to static relocation. To address these issues, we propose InstructUDrag, a diffusion-based framework that combines text instructions with object dragging, enabling simultaneous object dragging and text-based image editing. Our framework treats object dragging as an image reconstruction process, divided into two synergistic branches. The moving-reconstruction branch utilizes energy-based gradient guidance to move objects accurately, refining cross-attention maps to enhance relocation precision. The text-driven editing branch shares gradient signals with the reconstruction branch, ensuring consistent transformations and allowing fine-grained control over object attributes. We also employ DDPM inversion and inject prior information into noise maps to preserve the structure of moved objects. Extensive experiments demonstrate that InstructUDrag facilitates flexible, high-fidelity image editing, offering both precision in object relocation and semantic control over image content.

---

### Dual-granularity Sinkhorn Distillation for Enhanced Learning from Long-tailed Noisy Data

Authors: Feng Hong, Yu Huang, Zihua Zhao, Zhihan Zhou, Jiangchao Yao, Dongsheng Li, Ya Zhang, Yanfeng Wang
Published: 2025-10-09
Categories: cs.LG, cs.CV
arXiv: https://arxiv.org/abs/2510.08179v1

Real-world datasets for deep learning frequently suffer from the co-occurring challenges of class imbalance and label noise, hindering model performance. While methods exist for each issue, effectively combining them is non-trivial, as distinguishing genuine tail samples from noisy data proves difficult, often leading to conflicting optimization strategies. This paper presents a novel perspective: instead of primarily developing new complex techniques from scratch, we explore synergistically leveraging well-established, individually 'weak' auxiliary models - specialized for tackling either class imbalance or label noise but not both. This view is motivated by the insight that class imbalance (a distributional-level concern) and label noise (a sample-level concern) operate at different granularities, suggesting that robustness mechanisms for each can in principle offer complementary strengths without conflict. We propose Dual-granularity Sinkhorn Distillation (D-SINK), a novel framework that enhances dual robustness by distilling and integrating complementary insights from such 'weak', single-purpose auxiliary models. Specifically, D-SINK uses an optimal transport-optimized surrogate label allocation to align the target model's sample-level predictions with a noise-robust auxiliary and its class distributions with an imbalance-robust one. Extensive experiments on benchmark datasets demonstrate that D-SINK significantly improves robustness and achieves strong empirical performance in learning from long-tailed noisy data.

---

### Robust Canonicalization through Bootstrapped Data Re-Alignment

Authors: Johann Schmidt, Sebastian Stober
Published: 2025-10-09
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2510.08178v1

Fine-grained visual classification (FGVC) tasks, such as insect and bird identification, demand sensitivity to subtle visual cues while remaining robust to spatial transformations. A key challenge is handling geometric biases and noise, such as different orientations and scales of objects. Existing remedies rely on heavy data augmentation, which demands powerful models, or on equivariant architectures, which constrain expressivity and add cost. Canonicalization offers an alternative by shielding such biases from the downstream model. In practice, such functions are often obtained using canonicalization priors, which assume aligned training data. Unfortunately, real-world datasets never fulfill this assumption, causing the obtained canonicalizer to be brittle. We propose a bootstrapping algorithm that iteratively re-aligns training samples by progressively reducing variance and recovering the alignment assumption. We establish convergence guarantees under mild conditions for arbitrary compact groups, and show on four FGVC benchmarks that our method consistently outperforms equivariant, and canonicalization baselines while performing on par with augmentation.

---

### Long-tailed Recognition with Model Rebalancing

Authors: Jiaan Luo, Feng Hong, Qiang Hu, Xiaofeng Cao, Feng Liu, Jiangchao Yao
Published: 2025-10-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.08177v1

Long-tailed recognition is ubiquitous and challenging in deep learning and even in the downstream finetuning of foundation models, since the skew class distribution generally prevents the model generalization to the tail classes. Despite the promise of previous methods from the perspectives of data augmentation, loss rebalancing and decoupled training etc., consistent improvement in the broad scenarios like multi-label long-tailed recognition is difficult. In this study, we dive into the essential model capacity impact under long-tailed context, and propose a novel framework, Model Rebalancing (MORE), which mitigates imbalance by directly rebalancing the model's parameter space. Specifically, MORE introduces a low-rank parameter component to mediate the parameter space allocation guided by a tailored loss and sinusoidal reweighting schedule, but without increasing the overall model complexity or inference costs. Extensive experiments on diverse long-tailed benchmarks, spanning multi-class and multi-label tasks, demonstrate that MORE significantly improves generalization, particularly for tail classes, and effectively complements existing imbalance mitigation methods. These results highlight MORE's potential as a robust plug-and-play module in long-tailed settings.

---

### Leveraging Whisper Embeddings for Audio-based Lyrics Matching

Authors: Eleonora Mancini, Joan Serrà, Paolo Torroni, Yuki Mitsufuji
Published: 2025-10-09
Categories: cs.SD, cs.AI, cs.LG, eess.AS
arXiv: https://arxiv.org/abs/2510.08176v2

Audio-based lyrics matching can be an appealing alternative to other content-based retrieval approaches, but existing methods often suffer from limited reproducibility and inconsistent baselines. In this work, we introduce WEALY, a fully reproducible pipeline that leverages Whisper decoder embeddings for lyrics matching tasks. WEALY establishes robust and transparent baselines, while also exploring multimodal extensions that integrate textual and acoustic features. Through extensive experiments on standard datasets, we demonstrate that WEALY achieves a performance comparable to state-of-the-art methods that lack reproducibility. In addition, we provide ablation studies and analyses on language robustness, loss functions, and embedding strategies. This work contributes a reliable benchmark for future research, and underscores the potential of speech technologies for music information retrieval tasks.

---

### Prepared mind, fast response: A temporal decoupling framework for adaptive knowledge orchestration in open-domain dialogue

Authors: Jinling Gan, Churong Liang, Runnan Li
Published: 2025-10-09
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.08175v1

The latency-quality tradeoff is a fundamental constraint in open-domain dialogue AI systems, since comprehensive knowledge access necessitates prohibitive response delays. Contemporary approaches offer two inadequate solutions: lightweight instruct models achieve sub-second latency but lack reasoning depth, while tool-augmented ReAct agents enhance factuality through external knowledge at the cost of synchronous execution that blocks interaction during retrieval processes. PMFR is thus proposed, with a temporal decoupling framework that fundamentally resolves the contradiction through asynchronous knowledge orchestration. PMFR employs three coordinated components: (1) a Knowledge Adequacy Evaluator for real-time sufficiency assessment, (2) a Lightweight Response Generator for immediate user interaction, and (3) an Asynchronous Knowledge Refinement Agent for background knowledge enhancement. This architecture maintains continuous conversational flow while progressively enriching knowledge coverage through intelligent triggering mechanisms. Evaluation results on TopiOCQA demonstrate PMFR outperforms brute-force scaling: PMFR achieves 95.3% latency reduction (23.38s -> 1.09s) while preserving response quality comparable to heavyweight synchronous baselines (GEval-C: 0.613 vs. 0.620).

---

### NavSpace: How Navigation Agents Follow Spatial Intelligence Instructions

Authors: Haolin Yang, Yuxing Long, Zhuoyuan Yu, Zihan Yang, Minghan Wang, Jiapeng Xu, Yihan Wang, Ziyan Yu, Wenzhe Cai, Lei Kang, Hao Dong
Published: 2025-10-09
Categories: cs.RO, cs.AI, cs.CL, cs.CV
arXiv: https://arxiv.org/abs/2510.08173v2

Instruction-following navigation is a key step toward embodied intelligence. Prior benchmarks mainly focus on semantic understanding but overlook systematically evaluating navigation agents' spatial perception and reasoning capabilities. In this work, we introduce the NavSpace benchmark, which contains six task categories and 1,228 trajectory-instruction pairs designed to probe the spatial intelligence of navigation agents. On this benchmark, we comprehensively evaluate 22 navigation agents, including state-of-the-art navigation models and multimodal large language models. The evaluation results lift the veil on spatial intelligence in embodied navigation. Furthermore, we propose SNav, a new spatially intelligent navigation model. SNav outperforms existing navigation agents on NavSpace and real robot tests, establishing a strong baseline for future work.

---

### Bidirectional Representations Augmented Autoregressive Biological Sequence Generation

Authors: Xiang Zhang, Jiaqi Wei, Zijie Qiu, Sheng Xu, Zhi Jin, ZhiQiang Gao, Nanqing Dong, Siqi Sun
Published: 2025-10-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.08169v3

Autoregressive (AR) models, common in sequence generation, are limited in many biological tasks such as de novo peptide sequencing and protein modeling by their unidirectional nature, failing to capture crucial global bidirectional token dependencies. Non-Autoregressive (NAR) models offer holistic, bidirectional representations but face challenges with generative coherence and scalability. To transcend this, we propose a hybrid framework enhancing AR generation by dynamically integrating rich contextual information from non-autoregressive mechanisms. Our approach couples a shared input encoder with two decoders: a non-autoregressive one learning latent bidirectional biological features, and an AR decoder synthesizing the biological sequence by leveraging these bidirectional features. A novel cross-decoder attention module enables the AR decoder to iteratively query and integrate these bidirectional features, enriching its predictions. This synergy is cultivated via a tailored training strategy with importance annealing for balanced objectives and cross-decoder gradient blocking for stable, focused learning. Evaluations on a demanding nine-species benchmark of de novo peptide sequencing show that our model substantially surpasses AR and NAR baselines. It uniquely harmonizes AR stability with NAR contextual awareness, delivering robust, superior performance on diverse downstream data. This research advances biological sequence modeling techniques and contributes a novel architectural paradigm for augmenting AR models with enhanced bidirectional understanding for complex sequence generation. Code is available at https://github.com/BEAM-Labs/denovo.

---

### ARM2: Adaptive Reasoning Model with Vision Understanding and Executable Code

Authors: Jian Xie, Zhendong Chu, Aoxiao Zhong, Kai Zhang, Mingzhe Han, Xing Fan, Jialie Shen, Qingsong Wen
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.08163v3

Large Reasoning Models (LRMs) often suffer from the ``over-thinking'' problem, generating unnecessarily long reasoning on simple tasks. Some strategies have been proposed to mitigate this issue, such as length penalties or routing mechanisms, but they are typically heuristic and task-specific, lacking a general framework for adaptive reasoning. In this paper, we present ARM2, a unified model that adaptively balances reasoning performance and efficiency across multiple formats through a reinforcement learning framework augmented with length-aware optimization. Beyond conventional natural language inference, ARM2 integrates vision understanding, extending its applicability to multimodal. Moreover, ARM2 integrates executable code into reasoning, enabling substantial reductions in token cost while preserving task performance compared to long CoT. Experiments demonstrate that ARM2 achieves performance on par with traditional reasoning models trained with GRPO, while reducing token usage by over 70% on average. We further conduct extensive analyses to validate the effectiveness of ARM2 and the soundness of its design.

---

### Beyond Sub-6 GHz: Leveraging mmWave Wi-Fi for Gait-Based Person Identification

Authors: Nabeel Nisar Bhat, Maksim Karnaukh, Jakob Struye, Rafael Berkvens, Jeroen Famaey
Published: 2025-10-09
Categories: cs.LG, cs.HC
arXiv: https://arxiv.org/abs/2510.08160v1

Person identification plays a vital role in enabling intelligent, personalized, and secure human-computer interaction. Recent research has demonstrated the feasibility of leveraging Wi-Fi signals for passive person identification using a person's unique gait pattern. Although most existing work focuses on sub-6 GHz frequencies, the emergence of mmWave offers new opportunities through its finer spatial resolution, though its comparative advantages for person identification remain unexplored. This work presents the first comparative study between sub-6 GHz and mmWave Wi-Fi signals for person identification with commercial off-the-shelf (COTS) Wi-Fi, using a novel dataset of synchronized measurements from the two frequency bands in an indoor environment. To ensure a fair comparison, we apply identical training pipelines and model configurations across both frequency bands. Leveraging end-to-end deep learning, we show that even at low sampling rates (10 Hz), mmWave Wi-Fi signals can achieve high identification accuracy (91.2% on 20 individuals) when combined with effective background subtraction.

---

### Quantum Agents for Algorithmic Discovery

Authors: Iordanis Kerenidis, El-Amine Cherrat
Published: 2025-10-09
Categories: quant-ph, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2510.08159v1

We introduce quantum agents trained by episodic, reward-based reinforcement learning to autonomously rediscover several seminal quantum algorithms and protocols. In particular, our agents learn: efficient logarithmic-depth quantum circuits for the Quantum Fourier Transform; Grover's search algorithm; optimal cheating strategies for strong coin flipping; and optimal winning strategies for the CHSH and other nonlocal games. The agents achieve these results directly through interaction, without prior access to known optimal solutions. This demonstrates the potential of quantum intelligence as a tool for algorithmic discovery, opening the way for the automated design of novel quantum algorithms and protocols.

---

### Beyond Over-Refusal: Scenario-Based Diagnostics and Post-Hoc Mitigation for Exaggerated Refusals in LLMs

Authors: Shuzhou Yuan, Ercong Nie, Yinuo Sun, Chenxuan Zhao, William LaCroix, Michael Färber
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.08158v3

Large language models (LLMs) frequently produce false refusals, declining benign requests that contain terms resembling unsafe queries. We address this challenge by introducing two comprehensive benchmarks: the Exaggerated Safety Benchmark (XSB) for single-turn prompts, annotated with "Focus" keywords that identify refusal-inducing triggers, and the Multi-turn Scenario-based Exaggerated Safety Benchmark (MS-XSB), which systematically evaluates refusal calibration in realistic, context-rich dialog settings. Our benchmarks reveal that exaggerated refusals persist across diverse recent LLMs and are especially pronounced in complex, multi-turn scenarios. To mitigate these failures, we leverage post-hoc explanation methods to identify refusal triggers and deploy three lightweight, model-agnostic approaches, ignore-word instructions, prompt rephrasing, and attention steering, at inference time, all without retraining or parameter access. Experiments on four instruction-tuned Llama models demonstrate that these strategies substantially improve compliance on safe prompts while maintaining robust safety protections. Our findings establish a reproducible framework for diagnosing and mitigating exaggerated refusals, highlighting practical pathways to safer and more helpful LLM deployments.

---

### Beyond Textual CoT: Interleaved Text-Image Chains with Deep Confidence Reasoning for Image Editing

Authors: Zhentao Zou, Zhengrong Yue, Kunpeng Du, Binlei Bao, Hanting Li, Haizhen Xie, Guozheng Xu, Yue Zhou, Yali Wang, Jie Hu, Xue Jiang, Xinghao Chen
Published: 2025-10-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.08157v1

Image editing with natural language has gained significant popularity, yet existing methods struggle with intricate object intersections and fine-grained spatial relationships due to the lack of an explicit reasoning process. While Chain-of-Thought (CoT) has been explored to enhance reasoning, purely textual CoT or CoT augmented with coordinate information is fundamentally limited in its ability to represent intricate visual layouts and lacks the necessary visual cues to guide the generation of fine-grained, pixel-level details. To address these challenges, we propose Multimodal Reasoning Edit (MURE), a novel framework that shifts the visual editing process from purely text-based reasoning to a series of interleaved textual and visual rationales. Our framework performs image editing using a natively multimodal, interleaved text-image CoT. This approach generates a step-by-step chain of reasoning where a textual description is followed by a corresponding visual cue, such as a positional mask that defined intended edited regions or a representation of new content. Furthermore, to mitigate the hallucination phenomenon of large language models, we introduce Multimodal Deep Confidence (MMDC) reasoning paradigm. This paradigm explores a tree of visual reasoning paths at each step. By pruning low-quality branches using a deep confidence score from a reward model, it ensures the model consistently follows a high-quality trajectory towards the final edited result. The proposed method decomposes complex editing tasks into interdependent sub-tasks, achieving greater precision at each stage and yielding high-fidelity edited results. We define the formulation for interleaved text-image chains and release the first CoT-Edit-14K dataset, comprising 14K high-quality editing examples. Extensive experiments show that our method yields significant improvements across three image editing benchmarks.

---

### DACIP-RC: Domain Adaptive Continual Instruction Pre-Training via Reading Comprehension on Business Conversations

Authors: Elena Khasanova, Harsh Saini, Md Tahmid Rahman Laskar, Xue-Yong Fu, Cheng Chen, Shashi Bhushan TN
Published: 2025-10-09
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.08152v1

The rapid advancements in Large Language Models (LLMs) have enabled their adoption in real-world industrial scenarios for various natural language processing tasks. However, the high inference cost of large-scale LLMs makes their deployment impractical, necessitating the use of smaller models. Despite their efficiency, smaller LLMs lack robust zero-shot instruction-following capabilities across diverse domains, limiting their adaptability to dynamic user requirements. Traditional fine-tuning approaches exacerbate this issue by inducing catastrophic forgetting, reducing the model's generalization ability for unseen tasks. In this paper, we propose Domain Adaptive Continual Instruction Pre-Training via Reading Comprehension (DACIP-RC), a continual pre-training technique that enhances smaller LLMs' domain adaptability for business conversational tasks. Unlike conventional pre-training approaches that rely on next-token prediction, DACIP-RC generates diverse task instructions and responses via reading comprehension on conversation transcripts, enabling better instruction generalization. Our empirical evaluations demonstrate that DACIP-RC significantly improves zero-shot generalization across a wide range of business conversational tasks, including meeting summarization, action item generation, and call purpose identification. To the best of our knowledge, this is the first work to apply instruction pre-training on business conversational data, providing insights into how industries can leverage proprietary datasets for domain adaptation.

---

### Scaling Unsupervised Multi-Source Federated Domain Adaptation through Group-Wise Discrepancy Minimization

Authors: Larissa Reichart, Cem Ata Baykara, Ali Burak Ünal, Harlin Lee, Mete Akgün
Published: 2025-10-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.08150v3

Unsupervised multi-source domain adaptation (UMDA) leverages labeled data from multiple source domains to generalize to an unlabeled target. While federated UMDA addresses privacy by avoiding raw data sharing, existing methods scale poorly as the number of sources increases, often suffering from high computational overhead or training instability. We propose GALA, a scalable and robust federated UMDA framework designed for high-diversity settings. GALA achieves scalability by coupling a novel inter-group discrepancy minimization objective that approximates pairwise alignment with linear complexity alongside a temperature-controlled, centroid-based weighting strategy for dynamic source prioritization. These components enable stable, parallelizable training across many heterogeneous sources, addressing a critical scalability bottleneck that remains largely unaddressed in current literature. To evaluate performance in high-diversity scenarios, we introduce Digit-18, a new benchmark comprising 18 datasets with varied synthetic and real-world domain shifts. Extensive experiments demonstrate that GALA achieves state-of-the-art results on standard benchmarks and significantly outperforms prior methods in large-scale settings where others either fail to converge or become computationally infeasible.

---

### AI Knowledge Assist: An Automated Approach for the Creation of Knowledge Bases for Conversational AI Agents

Authors: Md Tahmid Rahman Laskar, Julien Bouvier Tremblay, Xue-Yong Fu, Cheng Chen, Shashi Bhushan TN
Published: 2025-10-09
Categories: cs.CL, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2510.08149v1

The utilization of conversational AI systems by leveraging Retrieval Augmented Generation (RAG) techniques to solve customer problems has been on the rise with the rapid progress of Large Language Models (LLMs). However, the absence of a company-specific dedicated knowledge base is a major barrier to the integration of conversational AI systems in contact centers. To this end, we introduce AI Knowledge Assist, a system that extracts knowledge in the form of question-answer (QA) pairs from historical customer-agent conversations to automatically build a knowledge base. Fine-tuning a lightweight LLM on internal data demonstrates state-of-the-art performance, outperforming larger closed-source LLMs. More specifically, empirical evaluation on 20 companies demonstrates that the proposed AI Knowledge Assist system that leverages the LLaMA-3.1-8B model eliminates the cold-start gap in contact centers by achieving above 90% accuracy in answering information-seeking questions. This enables immediate deployment of RAG-powered chatbots.

---

### Think Just Enough: Sequence-Level Entropy as a Confidence Signal for LLM Reasoning

Authors: Aman Sharma, Paras Chopra
Published: 2025-10-09
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.08146v3

We introduce a simple, yet novel entropy-based framework to drive token efficiency in large language models during reasoning tasks. Our approach uses Shannon entropy from token-level logprobs as a confidence signal to enable early stopping, achieving 25-50% computational savings while maintaining task accuracy. Crucially, we demonstrate that entropy-based confidence calibration represents an emergent property of advanced post-training optimization present in modern reasoning models but notably absent in standard instruction-tuned and pre-trained models (Llama 3.3 70B). We show that the entropy threshold to stop reasoning varies from model to model but can be calculated easily in one shot using only a few examples from existing reasoning datasets. Our results indicate that advanced reasoning models often know that they've gotten a correct answer early on, and that this emergent confidence awareness can be exploited to save tokens and reduce latency. The framework demonstrates consistent performance across reasoning-optimized model families with 25-50% computational cost reduction while preserving accuracy, revealing that confidence mechanisms represent a distinguishing characteristic of modern post-trained reasoning systems versus their predecessors.

---

### Mitigating Judgment Preference Bias in Large Language Models through Group-Based Polling

Authors: Shuliang Liu, Zhipeng Xu, Zhenghao Liu, Yukun Yan, Minghe Yu, Yu Gu, Chong Chen, Huiyuan Xie, Ge Yu
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.08145v2

Large Language Models (LLMs) as automatic evaluators, commonly referred to as LLM-as-a-Judge, have also attracted growing attention. This approach plays a vital role in aligning LLMs with human judgments, providing accurate and reliable assessments. However, LLM-based judgment models often exhibit judgment preference bias during the evaluation phase, tending to favor responses generated by themselves, undermining the reliability of their judgments. This paper introduces the Group-Based Polling Optimization (Genii), an unsupervised multi-agent collaborative optimization framework that mitigates the inherent judgment preference bias of judgment models. Specifically, Genii integrates various LLM-based judgment models into a multi-agent system and simulates the interactive client-server polling mechanism to optimize each client agent unsupervisedly. Our experiments demonstrate that Genii outperforms supervised models trained on annotated judgment data, while requiring no human-labeled annotations. Genii consistently improves performance across different client agents during the polling, even when weaker models act as server agents. Further analysis reveals that Genii effectively mitigates judgment preference bias of LLM-based judgment models, demonstrating its effectiveness. All codes are available at https://github.com/NEUIR/Genii.

---

### UniMMVSR: A Unified Multi-Modal Framework for Cascaded Video Super-Resolution

Authors: Shian Du, Menghan Xia, Chang Liu, Quande Liu, Xintao Wang, Pengfei Wan, Xiangyang Ji
Published: 2025-10-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.08143v1

Cascaded video super-resolution has emerged as a promising technique for decoupling the computational burden associated with generating high-resolution videos using large foundation models. Existing studies, however, are largely confined to text-to-video tasks and fail to leverage additional generative conditions beyond text, which are crucial for ensuring fidelity in multi-modal video generation. We address this limitation by presenting UniMMVSR, the first unified generative video super-resolution framework to incorporate hybrid-modal conditions, including text, images, and videos. We conduct a comprehensive exploration of condition injection strategies, training schemes, and data mixture techniques within a latent video diffusion model. A key challenge was designing distinct data construction and condition utilization methods to enable the model to precisely utilize all condition types, given their varied correlations with the target video. Our experiments demonstrate that UniMMVSR significantly outperforms existing methods, producing videos with superior detail and a higher degree of conformity to multi-modal conditions. We also validate the feasibility of combining UniMMVSR with a base model to achieve multi-modal guided generation of 4K video, a feat previously unattainable with existing techniques.

---

### SCOPE-RL: Stable and Quantitative Control of Policy Entropy in RL Post-Training

Authors: Chen Wang, Zhaochun Li, Jionghao Bai, Hexuan Deng, Ge Lan, Yue Wang
Published: 2025-10-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.08141v7

Reinforcement learning (RL) is a key paradigm for post-training large language models (LLMs), but the widely used Group Relative Policy Optimization (GRPO) often suffers from entropy collapse: exploration quickly disappears, policies converge prematurely, and sample diversity declines, ultimately harming training effectiveness. Existing remedies, including entropy bonuses and clip-based methods, rarely keep entropy within a stable exploration regime and often introduce oscillatory entropy or reward degradation. In this work, we identify a previously overlooked asymmetry in entropy dynamics: under high-temperature sampling, positive and negative samples have opposite effects on policy entropy. Specifically, high-temperature positive samples promote entropy growth, whereas negative samples suppress it. We provide a theoretical explanation for this phenomenon: when entropy decreases during policy updates, its derivative with respect to temperature is strictly positive under positive-sample updates, indicating that high-temperature positive samples can counteract entropy decay, thereby slowing entropy collapse and potentially reversing it. Motivated by this insight, we propose SCOPE-RL, a stable and quantitative entropy control framework through a regularization term constructed from temperature-adaptive positive samples. Extensive experiments show that SCOPE-RL consistently outperforms strong RL baselines on both Pass@1 and Pass@$k$. Our results provide evidence that escaping entropy collapse can improve reasoning performance, while also showing that the benefit is non-monotonic, with an optimal level of exploration for RL post-training in reasoning LLMs.

---

### Understanding Temporal Logic Consistency in Video-Language Models through Cross-Modal Attention Discriminability

Authors: Chengzhi Li, Heyan Huang, Ping Jian, Zhen Yang, Yaning Tian, Zhongbin Guo
Published: 2025-10-09
Categories: cs.CV, cs.AI, cs.MM
arXiv: https://arxiv.org/abs/2510.08138v2

Large language models (LLMs) often generate self-contradictory outputs, which severely impacts their reliability and hinders their adoption in practical applications. In video-language models (Video-LLMs), this phenomenon recently draws the attention of researchers. Specifically, these models fail to provide logically consistent responses to rephrased questions based on their grounding outputs. However, the underlying causes of this phenomenon remain underexplored. In this work, we adopt an interpretability-driven approach to analyze, statistically summarize, and intervention the potential factors of the phenomenon. We find that one of the primary reasons for the inconsistency in responses lies in the inability of cross-modal attention heads to effectively distinguish video tokens across different timestamps. To address this, we propose an attention enhancement method called Temporally Conditioned Attention Sharpening (TCAS), which constructs an enhancement objective based on attention distinctions to enhance the model's temporal resolution capability, thereby improving its temporal understanding logic consistency. Experimental results demonstrate that our method significantly enhances the temporal logic consistency of Video-LLMs. Further analyses reveal that our method indeed improves the temporal discriminability of attention heads, validating our conclusions. Additionally, our method even achieves performance improvements in general video temporal grounding tasks, suggesting that temporal logic consistency is an important factor in temporal understanding.

---

### Approximate Domain Unlearning for Vision-Language Models

Authors: Kodai Kawamura, Yuta Goto, Rintaro Yanagi, Hirokatsu Kataoka, Go Irie
Published: 2025-10-09
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.08132v1

Pre-trained Vision-Language Models (VLMs) exhibit strong generalization capabilities, enabling them to recognize a wide range of objects across diverse domains without additional training. However, they often retain irrelevant information beyond the requirements of specific downstream tasks, raising concerns about computational efficiency and potential information leakage. This has motivated growing interest in approximate unlearning, which aims to selectively remove unnecessary knowledge while preserving overall model performance. Existing approaches to approximate unlearning have primarily focused on class unlearning, where a VLM is retrained to fail to recognize specified object classes while maintaining accuracy for others. However, merely forgetting object classes is often insufficient in practical applications. For instance, an autonomous driving system should accurately recognize real cars while avoiding misrecognition of illustrated cars depicted in roadside advertisements as real cars, which could be hazardous. In this paper, we introduce Approximate Domain Unlearning (ADU), a novel problem setting that requires reducing recognition accuracy for images from specified domains (e.g., illustration) while preserving accuracy for other domains (e.g., real). ADU presents new technical challenges: due to the strong domain generalization capability of pre-trained VLMs, domain distributions are highly entangled in the feature space, making naive approaches based on penalizing target domains ineffective. To tackle this limitation, we propose a novel approach that explicitly disentangles domain distributions and adaptively captures instance-specific domain information. Extensive experiments show that our approach outperforms baselines built upon VLM tuning techniques, paving the way for practical and fine-grained unlearning in VLMs. Code: https://kodaikawamura.github.io/Domain_Unlearning/.

---

### Real-Time Motion-Controllable Autoregressive Video Diffusion

Authors: Kesen Zhao, Jiaxin Shi, Beier Zhu, Junbao Zhou, Xiaolong Shen, Yuan Zhou, Qianru Sun, Hanwang Zhang
Published: 2025-10-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.08131v3

Real-time motion-controllable video generation remains challenging due to the inherent latency of bidirectional diffusion models and the lack of effective autoregressive (AR) approaches. Existing AR video diffusion models are limited to simple control signals or text-to-video generation, and often suffer from quality degradation and motion artifacts in few-step generation. To address these challenges, we propose AR-Drag, the first RL-enhanced few-step AR video diffusion model for real-time image-to-video generation with diverse motion control. We first fine-tune a base I2V model to support basic motion control, then further improve it via reinforcement learning with a trajectory-based reward model. Our design preserves the Markov property through a Self-Rollout mechanism and accelerates training by selectively introducing stochasticity in denoising steps. Extensive experiments demonstrate that AR-Drag achieves high visual fidelity and precise motion alignment, significantly reducing latency compared with state-of-the-art motion-controllable VDMs, while using only 1.3B parameters. Additional visualizations can be found on our project page: https://kesenzhao.github.io/AR-Drag.github.io/.

---

### High-dimensional Analysis of Synthetic Data Selection

Authors: Parham Rezaei, Filip Kovacevic, Francesco Locatello, Marco Mondelli
Published: 2025-10-09
Categories: stat.ML, cs.LG
arXiv: https://arxiv.org/abs/2510.08123v1

Despite the progress in the development of generative models, their usefulness in creating synthetic data that improve prediction performance of classifiers has been put into question. Besides heuristic principles such as "synthetic data should be close to the real data distribution", it is actually not clear which specific properties affect the generalization error. Our paper addresses this question through the lens of high-dimensional regression. Theoretically, we show that, for linear models, the covariance shift between the target distribution and the distribution of the synthetic data affects the generalization error but, surprisingly, the mean shift does not. Furthermore we prove that, in some settings, matching the covariance of the target distribution is optimal. Remarkably, the theoretical insights from linear models carry over to deep neural networks and generative models. We empirically demonstrate that the covariance matching procedure (matching the covariance of the synthetic data with that of the data coming from the target distribution) performs well against several recent approaches for synthetic data selection, across training paradigms, architectures, datasets and generative models used for augmentation.

---

### Interpreting LLM-as-a-Judge Policies via Verifiable Global Explanations

Authors: Jasmina Gajcin, Erik Miehling, Rahul Nair, Elizabeth Daly, Radu Marinescu, Seshu Tirupathi
Published: 2025-10-09
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.08120v1

Using LLMs to evaluate text, that is, LLM-as-a-judge, is increasingly being used at scale to augment or even replace human annotations. As such, it is imperative that we understand the potential biases and risks of doing so. In this work, we propose an approach for extracting high-level concept-based global policies from LLM-as-a-Judge. Our approach consists of two algorithms: 1) CLoVE (Contrastive Local Verifiable Explanations), which generates verifiable, concept-based, contrastive local explanations and 2) GloVE (Global Verifiable Explanations), which uses iterative clustering, summarization and verification to condense local rules into a global policy. We evaluate GloVE on seven standard benchmarking datasets for content harm detection. We find that the extracted global policies are highly faithful to decisions of the LLM-as-a-Judge. Additionally, we evaluated the robustness of global policies to text perturbations and adversarial attacks. Finally, we conducted a user study to evaluate user understanding and satisfaction with global policies.

---

### Near-optimal Rank Adaptive Inference of High Dimensional Matrices

Authors: Frédéric Zheng, Yassir Jedra, Alexandre Proutiere
Published: 2025-10-09
Categories: cs.IT, stat.ML
arXiv: https://arxiv.org/abs/2510.08117v2

We address the problem of estimating a high-dimensional matrix from linear measurements, with a focus on designing optimal rank-adaptive algorithms. These algorithms infer the matrix by estimating its singular values and the corresponding singular vectors up to an effective rank, adaptively determined based on the data. We establish instance-specific lower bounds for the sample complexity of such algorithms, uncovering fundamental trade-offs in selecting the effective rank: balancing the precision of estimating a subset of singular values against the approximation cost incurred for the remaining ones. Our analysis identifies how the optimal effective rank depends on the matrix being estimated, the sample size, and the noise level. We propose an algorithm that combines a Least-Squares estimator with a universal singular value thresholding procedure. We provide finite-sample error bounds for this algorithm and demonstrate that its performance nearly matches the derived fundamental limits. Our results rely on an enhanced analysis of matrix denoising methods based on singular value thresholding. We validate our findings with applications to multivariate regression and linear dynamical system identification.

---

### Random Window Augmentations for Deep Learning Robustness in CT and Liver Tumor Segmentation

Authors: Eirik A. Østmo, Kristoffer K. Wickstrøm, Keyur Radiya, Michael C. Kampffmeyer, Karl Øyvind Mikalsen, Robert Jenssen
Published: 2025-10-09
Categories: cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2510.08116v1

Contrast-enhanced Computed Tomography (CT) is important for diagnosis and treatment planning for various medical conditions. Deep learning (DL) based segmentation models may enable automated medical image analysis for detecting and delineating tumors in CT images, thereby reducing clinicians' workload. Achieving generalization capabilities in limited data domains, such as radiology, requires modern DL models to be trained with image augmentation. However, naively applying augmentation methods developed for natural images to CT scans often disregards the nature of the CT modality, where the intensities measure Hounsfield Units (HU) and have important physical meaning. This paper challenges the use of such intensity augmentations for CT imaging and shows that they may lead to artifacts and poor generalization. To mitigate this, we propose a CT-specific augmentation technique, called Random windowing, that exploits the available HU distribution of intensities in CT images. Random windowing encourages robustness to contrast-enhancement and significantly increases model performance on challenging images with poor contrast or timing. We perform ablations and analysis of our method on multiple datasets, and compare to, and outperform, state-of-the-art alternatives, while focusing on the challenge of liver tumor segmentation.

---

### Can Risk-taking AI-Assistants suitably represent entities

Authors: Ali Mazyaki, Mohammad Naghizadeh, Samaneh Ranjkhah Zonouzaghi, Amirhossein Farshi Sotoudeh
Published: 2025-10-09
Categories: cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2510.08114v1

Responsible AI demands systems whose behavioral tendencies can be effectively measured, audited, and adjusted to prevent inadvertently nudging users toward risky decisions or embedding hidden biases in risk aversion. As language models (LMs) are increasingly incorporated into AI-driven decision support systems, understanding their risk behaviors is crucial for their responsible deployment. This study investigates the manipulability of risk aversion (MoRA) in LMs, examining their ability to replicate human risk preferences across diverse economic scenarios, with a focus on gender-specific attitudes, uncertainty, role-based decision-making, and the manipulability of risk aversion. The results indicate that while LMs such as DeepSeek Reasoner and Gemini-2.0-flash-lite exhibit some alignment with human behaviors, notable discrepancies highlight the need to refine bio-centric measures of manipulability. These findings suggest directions for refining AI design to better align human and AI risk preferences and enhance ethical decision-making. The study calls for further advancements in model design to ensure that AI systems more accurately replicate human risk preferences, thereby improving their effectiveness in risk management contexts. This approach could enhance the applicability of AI assistants in managing risk.

---

### Bayesian Decision Making around Experts

Authors: Daniel Jarne Ornia, Joel Dyer, Nicholas Bishop, Anisoara Calinescu, Michael Wooldridge
Published: 2025-10-09
Categories: cs.LG, cs.AI, cs.MA
arXiv: https://arxiv.org/abs/2510.08113v1

Complex learning agents are increasingly deployed alongside existing experts, such as human operators or previously trained agents. However, it remains unclear how should learners optimally incorporate certain forms of expert data, which may differ in structure from the learner's own action-outcome experiences. We study this problem in the context of Bayesian multi-armed bandits, considering: (i) offline settings, where the learner receives a dataset of outcomes from the expert's optimal policy before interaction, and (ii) simultaneous settings, where the learner must choose at each step whether to update its beliefs based on its own experience, or based on the outcome simultaneously achieved by an expert. We formalize how expert data influences the learner's posterior, and prove that pretraining on expert outcomes tightens information-theoretic regret bounds by the mutual information between the expert data and the optimal action. For the simultaneous setting, we propose an information-directed rule where the learner processes the data source that maximizes their one-step information gain about the optimal action. Finally, we propose strategies for how the learner can infer when to trust the expert and when not to, safeguarding the learner for the cases where the expert is ineffective or compromised. By quantifying the value of expert data, our framework provides practical, information-theoretic algorithms for agents to intelligently decide when to learn from others.

---

### Evaluating LLM-Generated Legal Explanations for Regulatory Compliance in Social Media Influencer Marketing

Authors: Haoyang Gui, Thales Bertaglia, Taylor Annabell, Catalina Goanta, Tjomme Dooper, Gerasimos Spanakis
Published: 2025-10-09
Categories: cs.CL, cs.CY
arXiv: https://arxiv.org/abs/2510.08111v1

The rise of influencer marketing has blurred boundaries between organic content and sponsored content, making the enforcement of legal rules relating to transparency challenging. Effective regulation requires applying legal knowledge with a clear purpose and reason, yet current detection methods of undisclosed sponsored content generally lack legal grounding or operate as opaque "black boxes". Using 1,143 Instagram posts, we compare gpt-5-nano and gemini-2.5-flash-lite under three prompting strategies with controlled levels of legal knowledge provided. Both models perform strongly in classifying content as sponsored or not (F1 up to 0.93), though performance drops by over 10 points on ambiguous cases. We further develop a taxonomy of reasoning errors, showing frequent citation omissions (28.57%), unclear references (20.71%), and hidden ads exhibiting the highest miscue rate (28.57%). While adding regulatory text to the prompt improves explanation quality, it does not consistently improve detection accuracy. The contribution of this paper is threefold. First, it makes a novel addition to regulatory compliance technology by providing a taxonomy of common errors in LLM-generated legal reasoning to evaluate whether automated moderation is not only accurate but also legally robust, thereby advancing the transparent detection of influencer marketing content. Second, it features an original dataset of LLM explanations annotated by two students who were trained in influencer marketing law. Third, it combines quantitative and qualitative evaluation strategies for LLM explanations and critically reflects on how these findings can support advertising regulatory bodies in automating moderation processes on a solid legal foundation.

---