# Research Papers — 2025-10 · bundle page 157 of 217

> Full-text markdown of papers #7801–7850 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/157
[← Previous bundle](https://aipod.dev/papers/archive/2025-10/page/156.md)
[Next bundle →](https://aipod.dev/papers/archive/2025-10/page/158.md)

### Fewer Weights, More Problems: A Practical Attack on LLM Pruning

Authors: Kazuki Egashira, Robin Staab, Thibaud Gloaguen, Mark Vero, Martin Vechev
Published: 2025-10-09
Categories: cs.LG, cs.AI, cs.CR
arXiv: https://arxiv.org/abs/2510.07985v3

Model pruning, i.e., removing a subset of model weights, has become a prominent approach to reducing the memory footprint of large language models (LLMs) during inference. Notably, popular inference engines, such as vLLM, enable users to conveniently prune downloaded models before they are deployed. While the utility and efficiency of pruning methods have improved significantly, the security implications of pruning remain underexplored. In this work, for the first time, we show that modern LLM pruning methods can be maliciously exploited. In particular, an adversary can construct a model that appears benign yet, once pruned, exhibits malicious behaviors. Our method is based on the idea that the adversary can compute a proxy metric that estimates how likely each parameter is to be pruned. With this information, the adversary can first inject a malicious behavior into those parameters that are unlikely to be pruned. Then, they can repair the model by using parameters that are likely to be pruned, effectively canceling out the injected behavior in the unpruned model. We demonstrate the severity of our attack through extensive evaluation on five models; after any of the pruning in vLLM are applied (Magnitude, Wanda, and SparseGPT), it consistently exhibits strong malicious behaviors in a diverse set of attack scenarios (success rates of up to $95.7\%$ for jailbreak, $98.7\%$ for benign instruction refusal, and $99.5\%$ for targeted content injection). Our results reveal a critical deployment-time security gap and underscore the urgent need for stronger security awareness in model compression.

---

### Is Architectural Complexity Always the Answer? A Case Study on SwinIR vs. an Efficient CNN

Authors: Chandresh Sutariya, Nitin Singh
Published: 2025-10-09
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2510.07984v1

The simultaneous restoration of high-frequency details and suppression of severe noise in low-light imagery presents a significant and persistent challenge in computer vision. While large-scale Transformer models like SwinIR have set the state of the art in performance, their high computational cost can be a barrier for practical applications. This paper investigates the critical trade-off between performance and efficiency by comparing the state-of-the-art SwinIR model against a standard, lightweight Convolutional Neural Network (CNN) on this challenging task. Our experimental results reveal a nuanced but important finding. While the Transformer-based SwinIR model achieves a higher peak performance, with a Peak Signal-to-Noise Ratio (PSNR) of 39.03 dB, the lightweight CNN delivers a surprisingly competitive PSNR of 37.4 dB. Crucially, the CNN reached this performance after converging in only 10 epochs of training, whereas the more complex SwinIR model required 132 epochs. This efficiency is further underscored by the model's size; the CNN is over 55 times smaller than SwinIR. This work demonstrates that a standard CNN can provide a near state-of-the-art result with significantly lower computational overhead, presenting a compelling case for its use in real-world scenarios where resource constraints are a primary concern.

---

### ZeroCard: Cardinality Estimation with Zero Dependence on Target Databases -- No Data, No Query, No Retraining

Authors: Xianghong Xu, Rong Kang, Xiao He, Lei Zhang, Jianjun Chen, Tieying Zhang
Published: 2025-10-09
Categories: cs.DB, cs.AI
arXiv: https://arxiv.org/abs/2510.07983v1

Cardinality estimation is a fundamental task in database systems and plays a critical role in query optimization. Despite significant advances in learning-based cardinality estimation methods, most existing approaches remain difficult to generalize to new datasets due to their strong dependence on raw data or queries, thus limiting their practicality in real scenarios. To overcome these challenges, we argue that semantics in the schema may benefit cardinality estimation, and leveraging such semantics may alleviate these dependencies. To this end, we introduce ZeroCard, the first semantics-driven cardinality estimation method that can be applied without any dependence on raw data access, query logs, or retraining on the target database. Specifically, we propose to predict data distributions using schema semantics, thereby avoiding raw data dependence. Then, we introduce a query template-agnostic representation method to alleviate query dependence. Finally, we construct a large-scale query dataset derived from real-world tables and pretrain ZeroCard on it, enabling it to learn cardinality from schema semantics and predicate representations. After pretraining, ZeroCard's parameters can be frozen and applied in an off-the-shelf manner. We conduct extensive experiments to demonstrate the distinct advantages of ZeroCard and show its practical applications in query optimization. Its zero-dependence property significantly facilitates deployment in real-world scenarios.

---

### Unveiling the Power of Multiple Gossip Steps: A Stability-Based Generalization Analysis in Decentralized Training

Authors: Qinglun Li, Yingqi Liu, Miao Zhang, Xiaochun Cao, Quanjun Yin, Li Shen
Published: 2025-10-09
Categories: cs.LG, cs.AI, math.NA
arXiv: https://arxiv.org/abs/2510.07980v1

Decentralized training removes the centralized server, making it a communication-efficient approach that can significantly improve training efficiency, but it often suffers from degraded performance compared to centralized training. Multi-Gossip Steps (MGS) serve as a simple yet effective bridge between decentralized and centralized training, significantly reducing experiment performance gaps. However, the theoretical reasons for its effectiveness and whether this gap can be fully eliminated by MGS remain open questions. In this paper, we derive upper bounds on the generalization error and excess error of MGS using stability analysis, systematically answering these two key questions. 1). Optimization Error Reduction: MGS reduces the optimization error bound at an exponential rate, thereby exponentially tightening the generalization error bound and enabling convergence to better solutions. 2). Gap to Centralization: Even as MGS approaches infinity, a non-negligible gap in generalization error remains compared to centralized mini-batch SGD ($\mathcal{O}(T^{\frac{cβ}{cβ+1}}/{n m})$ in centralized and $\mathcal{O}(T^{\frac{2cβ}{2cβ+2}}/{n m^{\frac{1}{2cβ+2}}})$ in decentralized). Furthermore, we provide the first unified analysis of how factors like learning rate, data heterogeneity, node count, per-node sample size, and communication topology impact the generalization of MGS under non-convex settings without the bounded gradients assumption, filling a critical theoretical gap in decentralized training. Finally, promising experiments on CIFAR datasets support our theoretical findings.

---

### VoiceAgentBench: Are Voice Assistants ready for agentic tasks?

Authors: Dhruv Jain, Harshit Shukla, Gautam Rajeev, Ashish Kulkarni, Chandra Khatri, Shubham Agarwal
Published: 2025-10-09
Categories: cs.AI, cs.CL, cs.LG
arXiv: https://arxiv.org/abs/2510.07978v3

Large scale Speech Language Models have enabled voice assistants capable of understanding natural spoken queries and performing complex tasks. However, existing speech benchmarks largely focus on isolated capabilities such as transcription or question answering and do not systematically evaluate agentic behavior or adversarial robustness. To address this, we introduce VoiceAgentBench, a comprehensive benchmark for evaluating SpeechLMs in realistic spoken agentic settings, comprising 6,000+ synthetic spoken queries spanning single-tool invocations, multi-tool workflows, multi-turn dialogue, and safety evaluations across English and six Indic languages. To ensure speaker diversity, we further simulate speaker variability using a novel sampling strategy that selects audios for TTS voice conversion based on speaker embeddings to maximize acoustic diversity. Our evaluation measures tool selection accuracy, structural consistency, and the correctness of tool invocations, including adversarial robustness. Across agentic tasks, ASR-LLM pipelines outperform end-to-end SpeechLMs, achieving up to 60.6% average parameter-filling accuracy on English, while SpeechLMs exhibit lower performance and sharper degradation on Indic languages. All models struggle in sequential workflows and safety evaluations, highlighting persistent limitations in tool orchestration, multilingual generalization, and safety robustness. VoiceAgentBench is publicly available on Hugging Face at https://huggingface.co/datasets/krutrim-ai-labs/VoiceAgentBench, and the codebase is released at https://github.com/ola-krutrim/VoiceAgentBench.

---

### The impact of abstract and object tags on image privacy classification

Authors: Darya Baranouskaya, Andrea Cavallaro
Published: 2025-10-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.07976v2

Object tags denote concrete entities and are central to many computer vision tasks, whereas abstract tags capture higher-level information, which is relevant for tasks that require a contextual, potentially subjective scene understanding. Object and abstract tags extracted from images also facilitate interpretability. In this paper, we explore which type of tags is more suitable for the context-dependent and inherently subjective task of image privacy. While object tags are generally used for privacy classification, we show that abstract tags are more effective when the tag budget is limited. Conversely, when a larger number of tags per image is available, object-related information is as useful. We believe that these findings will guide future research in developing more accurate image privacy classifiers, informed by the role of tag types and quantity.

---

### Executable Analytic Concepts as the Missing Link Between VLM Insight and Precise Manipulation

Authors: Mingyang Sun, Jiude Wei, Qichen He, Donglin Wang, Cewu Lu, Jianhua Sun
Published: 2025-10-09
Categories: cs.RO, cs.AI
arXiv: https://arxiv.org/abs/2510.07975v1

Enabling robots to perform precise and generalized manipulation in unstructured environments remains a fundamental challenge in embodied AI. While Vision-Language Models (VLMs) have demonstrated remarkable capabilities in semantic reasoning and task planning, a significant gap persists between their high-level understanding and the precise physical execution required for real-world manipulation. To bridge this "semantic-to-physical" gap, we introduce GRACE, a novel framework that grounds VLM-based reasoning through executable analytic concepts (EAC)-mathematically defined blueprints that encode object affordances, geometric constraints, and semantics of manipulation. Our approach integrates a structured policy scaffolding pipeline that turn natural language instructions and visual information into an instantiated EAC, from which we derive grasp poses, force directions and plan physically feasible motion trajectory for robot execution. GRACE thus provides a unified and interpretable interface between high-level instruction understanding and low-level robot control, effectively enabling precise and generalizable manipulation through semantic-physical grounding. Extensive experiments demonstrate that GRACE achieves strong zero-shot generalization across a variety of articulated objects in both simulated and real-world environments, without requiring task-specific training.

---

### Active Confusion Expression in Large Language Models: Leveraging World Models toward Better Social Reasoning

Authors: Jialu Du, Guiyang Hou, Yihui Fu, Chen Wu, Wenqi Zhang, Yongliang Shen, Weiming Lu
Published: 2025-10-09
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.07974v2

While large language models (LLMs) excel in mathematical and code reasoning, we observe they struggle with social reasoning tasks, exhibiting cognitive confusion, logical inconsistencies, and conflation between objective world states and subjective belief states. Through deteiled analysis of DeepSeek-R1's reasoning trajectories, we find that LLMs frequently encounter reasoning impasses and tend to output contradictory terms like "tricky" and "confused" when processing scenarios with multiple participants and timelines, leading to erroneous reasoning or infinite loops. The core issue is their inability to disentangle objective reality from agents' subjective beliefs. To address this, we propose an adaptive world model-enhanced reasoning mechanism that constructs a dynamic textual world model to track entity states and temporal sequences. It dynamically monitors reasoning trajectories for confusion indicators and promptly intervenes by providing clear world state descriptions, helping models navigate through cognitive dilemmas. The mechanism mimics how humans use implicit world models to distinguish between external events and internal beliefs. Evaluations on three social benchmarks demonstrate significant improvements in accuracy (e.g., +10% in Hi-ToM) while reducing computational costs (up to 33.8% token reduction), offering a simple yet effective solution for deploying LLMs in social contexts.

---

### SHE: Stepwise Hybrid Examination Reinforcement Learning Framework for E-commerce Search Relevance

Authors: Pengkun Jiao, Yiming Jin, Jianhui Yang, Chenhe Dong, Zerui Huang, Shaowei Yao, Xiaojiang Zhou, Dan Ou, Haihong Tang
Published: 2025-10-09
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.07972v3

Query-product relevance prediction is vital for AI-driven e-commerce, yet current LLM-based approaches face a dilemma: SFT and DPO struggle with long-tail generalization due to coarse supervision, while traditional RLVR suffers from sparse feedback that fails to correct intermediate reasoning errors. We propose Stepwise Hybrid Examination (SHE), an RL framework that ensures logical consistency through Stepwise Reward Policy Optimization (SRPO). SRPO utilizes a hybrid reward mechanism-combining generative reward models with human-annotated verifiers-to provide fine-grained, step-level signals. To further enhance stability, SHE incorporates diversified data filtering to maintain policy entropy and a multi-stage curriculum learning protocol for progressive skill acquisition. Extensive experiments on real-world search benchmarks show that SHE improves both reasoning quality and relevance-prediction accuracy in large-scale e-commerce settings, outperforming SFT, DPO, GRPO, and other baselines, while also enhancing interpretability and robustness.

---

### Climate Surrogates for Scalable Multi-Agent Reinforcement Learning: A Case Study with CICERO-SCM

Authors: Oskar Bohn Lassen, Serio Angelo Maria Agriesti, Filipe Rodrigues, Blaz Kurnik, Francisco Camara Pereira
Published: 2025-10-09
Categories: cs.LG, cs.MA
arXiv: https://arxiv.org/abs/2510.07971v2

Climate policy analysis requires models that capture multi-gas climate effects, but such models are too slow to embed in reinforcement learning loops at scale. In collaboration with the European Environment Agency, we develop a multi-agent reinforcement learning (MARL) framework that integrates a higher-fidelity climate surrogate as the environment transition, enabling regional agents to learn policies under multi-gas dynamics. We train a recurrent surrogate on $20{,}000$ multi-gas emission pathways to emulate CICERO-SCM. The surrogate achieves near-simulator accuracy (global-mean temperature RMSE $\approx\!4\!\times \!10^{-4}\,\mathrm{K}$) with $\sim\!1000\times$ faster one-step inference and yields $>\!100\times$ end-to-end MARL training speed-up. We show policy agreement with the simulator in tractable settings and propose a replay- and rank-consistency test (Kendall's $τ$) for assessing policy fidelity when simulator-in-the-loop training is infeasible. This enables large-scale multi-agent policy experiments while retaining high-fidelity multi-gas climate response.

---

### Stick-Breaking Mixture Normalizing Flows with Component-Wise Tail Adaptation for Variational Inference

Authors: Seungsu Han, Juyoung Hwang, Won Chang
Published: 2025-10-09
Categories: stat.ML, cs.LG
arXiv: https://arxiv.org/abs/2510.07965v1

Normalizing flows with a Gaussian base provide a computationally efficient way to approximate posterior distributions in Bayesian inference, but they often struggle to capture complex posteriors with multimodality and heavy tails. We propose a stick-breaking mixture base with component-wise tail adaptation (StiCTAF) for posterior approximation. The method first learns a flexible mixture base to mitigate the mode-seeking bias of reverse KL divergence through a weighted average of component-wise ELBOs. It then estimates local tail indices of unnormalized densities and finally refines each mixture component using a shared backbone combined with component-specific tail transforms calibrated by the estimated indices. This design enables accurate mode coverage and anisotropic tail modeling while retaining exact density evaluation and stable optimization. Experiments on synthetic posteriors demonstrate improved tail recovery and better coverage of multiple modes compared to benchmark models. We also present a real-data analysis illustrating the practical benefits of our approach for posterior inference.

---

### PRESCRIBE: Predicting Single-Cell Responses with Bayesian Estimation

Authors: Jiabei Cheng, Changxi Chi, Jingbo Zhou, Hongyi Xin, Jun Xia
Published: 2025-10-09
Categories: cs.LG, q-bio.QM
arXiv: https://arxiv.org/abs/2510.07964v1

In single-cell perturbation prediction, a central task is to forecast the effects of perturbing a gene unseen in the training data. The efficacy of such predictions depends on two factors: (1) the similarity of the target gene to those covered in the training data, which informs model (epistemic) uncertainty, and (2) the quality of the corresponding training data, which reflects data (aleatoric) uncertainty. Both factors are critical for determining the reliability of a prediction, particularly as gene perturbation is an inherently stochastic biochemical process. In this paper, we propose PRESCRIBE (PREdicting Single-Cell Response wIth Bayesian Estimation), a multivariate deep evidential regression framework designed to measure both sources of uncertainty jointly. Our analysis demonstrates that PRESCRIBE effectively estimates a confidence score for each prediction, which strongly correlates with its empirical accuracy. This capability enables the filtering of untrustworthy results, and in our experiments, it achieves steady accuracy improvements of over 3% compared to comparable baselines.

---

### LightReasoner: Can Small Language Models Teach Large Language Models Reasoning?

Authors: Jingyuan Wang, Yankai Chen, Zhonghang Li, Chao Huang
Published: 2025-10-09
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.07962v2

Large language models (LLMs) have demonstrated remarkable progress in reasoning, often through supervised fine-tuning (SFT). However, SFT is resource-intensive, relying on large curated datasets, rejection-sampled demonstrations, and uniform optimization across all tokens, even though only a fraction carry meaningful learning value. In this work, we explore a counterintuitive idea: can smaller language models (SLMs) teach larger language models (LLMs) by revealing high-value reasoning moments that reflect the latter's unique strength? We propose LightReasoner, a novel framework that leverages the behavioral divergence between a stronger expert model (LLM) and a weaker amateur model (SLM). LightReasoner operates in two stages: (1) a sampling stage that pinpoints critical reasoning moments and constructs supervision examples capturing the expert's advantage through expert-amateur contrast, and (2) a fine-tuning stage that aligns the expert model with these distilled examples, amplifying its reasoning strengths. Across seven mathematical benchmarks, LightReasoner improves accuracy by up to 28.1%, while reducing time consumption by 90%, sampled problems by 80%, and tuned token usage by 99%, all without relying on ground-truth labels. By turning weaker SLMs into effective teaching signals, LightReasoner offers a scalable and resource-efficient approach for advancing LLM reasoning. Code is available at: https://github.com/HKUDS/LightReasoner

---

### Latent Harmony: Synergistic Unified UHD Image Restoration via Latent Space Regularization and Controllable Refinement

Authors: Yidi Liu, Xueyang Fu, Jie Huang, Jie Xiao, Dong Li, Wenlong Zhang, Lei Bai, Zheng-Jun Zha
Published: 2025-10-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.07961v3

Ultra-High Definition (UHD) image restoration faces a trade-off between computational efficiency and high-frequency detail retention. While Variational Autoencoders (VAEs) improve efficiency via latent-space processing, their Gaussian constraint often discards degradation-specific high-frequency information, hurting reconstruction fidelity. To overcome this, we propose Latent Harmony, a two-stage framework that redefines VAEs for UHD restoration by jointly regularizing the latent space and enforcing high-frequency-aware reconstruction.In Stage One, we introduce LH-VAE, which enhances semantic robustness through visual semantic constraints and progressive degradation perturbations, while latent equivariance strengthens high-frequency reconstruction.Stage Two jointly trains this refined VAE with a restoration model using High-Frequency Low-Rank Adaptation (HF-LoRA): an encoder LoRA guided by a fidelity-oriented high-frequency alignment loss to recover authentic details, and a decoder LoRA driven by a perception-oriented loss to synthesize realistic textures. Both LoRA modules are trained via alternating optimization with selective gradient propagation to preserve the pretrained latent structure.At inference, a tunable parameter α enables flexible fidelity-perception trade-offs.Experiments show Latent Harmony achieves state-of-the-art performance across UHD and standard-resolution tasks, effectively balancing efficiency, perceptual quality, and reconstruction accuracy.

---

### A Systematic Evaluation of Self-Supervised Learning for Label-Efficient Sleep Staging with Wearable EEG

Authors: Emilio Estevan, María Sierra-Torralba, Eduardo López-Larraz, Luis Montesano
Published: 2025-10-09
Categories: cs.HC, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2510.07960v3

Wearable EEG devices have emerged as a promising alternative to polysomnography (PSG). As affordable and scalable solutions, their widespread adoption results in the collection of massive volumes of unlabeled data that cannot be analyzed by clinicians at scale. Meanwhile, the recent success of deep learning for sleep scoring has relied on large annotated datasets. Self-supervised learning (SSL) offers an opportunity to bridge this gap, leveraging unlabeled signals to address label scarcity and reduce annotation effort. In this paper, we present the first systematic evaluation of SSL for sleep staging using wearable EEG. We introduce a structured benchmarking framework encompassing a range of SSL paradigms and propose a specialized pipeline tailored to the wearable EEG domain, evaluating them on two sleep databases acquired with the Ikon Sleep wearable headband: BOAS, a high-quality benchmark containing PSG and wearable EEG recordings with consensus labels, and HOGAR, a large collection of home-based, self-recorded, and unlabeled recordings. Three evaluation scenarios are defined to study label efficiency, representation quality, and cross-dataset generalization. Results show that SSL consistently improves classification performance by up to 10% over supervised baselines, with gains particularly evident when labeled data is scarce. SSL achieves clinical-grade accuracy above 80% leveraging only 5% to 10% of labeled data, while the supervised approach requires twice the labels. Additionally, the proposed domain-specific SSL pipeline outperforms the evaluated general-purpose EEG foundation models across all scenarios. Our findings demonstrate the potential of SSL to enable label-efficient sleep staging with wearable EEG, reducing reliance on manual annotations and advancing the development of affordable sleep monitoring systems.

---

### DISCO: Diversifying Sample Condensation for Efficient Model Evaluation

Authors: Alexander Rubinstein, Benjamin Raible, Martin Gubri, Seong Joon Oh
Published: 2025-10-09
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.07959v2

Evaluating modern machine learning models has become prohibitively expensive. Benchmarks such as LMMs-Eval and HELM demand thousands of GPU hours per model. Costly evaluation reduces inclusivity, slows the cycle of innovation, and worsens environmental impact. The typical approach follows two steps. First, select an anchor subset of data. Second, train a mapping from the accuracy on this subset to the final test result. The drawback is that anchor selection depends on clustering, which can be complex and sensitive to design choices. We argue that promoting diversity among samples is not essential; what matters is to select samples that $\textit{maximise diversity in model responses}$. Our method, $\textbf{Diversifying Sample Condensation (DISCO)}$, selects the top-k samples with the greatest model disagreements. This uses greedy, sample-wise statistics rather than global clustering. The approach is conceptually simpler. From a theoretical view, inter-model disagreement provides an information-theoretically optimal rule for such greedy selection. $\textbf{DISCO}$ shows empirical gains over prior methods, achieving state-of-the-art results in performance prediction across MMLU, Hellaswag, Winogrande, and ARC. Code is available here: https://github.com/arubique/disco-public.

---

### A$^2$Search: Ambiguity-Aware Question Answering with Reinforcement Learning

Authors: Fengji Zhang, Xinyao Niu, Chengyang Ying, Guancheng Lin, Zhongkai Hao, Zhou Fan, Chengen Huang, Jacky Keung, Bei Chen, Junyang Lin
Published: 2025-10-09
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.07958v1

Recent advances in Large Language Models (LLMs) and Reinforcement Learning (RL) have led to strong performance in open-domain question answering (QA). However, existing models still struggle with questions that admit multiple valid answers. Standard QA benchmarks, which typically assume a single gold answer, overlook this reality and thus produce inappropriate training signals. Existing attempts to handle ambiguity often rely on costly manual annotation, which is difficult to scale to multi-hop datasets such as HotpotQA and MuSiQue. In this paper, we present A$^2$Search, an annotation-free, end-to-end training framework to recognize and handle ambiguity. At its core is an automated pipeline that detects ambiguous questions and gathers alternative answers via trajectory sampling and evidence verification. The model is then optimized with RL using a carefully designed $\mathrm{AnsF1}$ reward, which naturally accommodates multiple answers. Experiments on eight open-domain QA benchmarks demonstrate that A$^2$Search achieves new state-of-the-art performance. With only a single rollout, A$^2$Search-7B yields an average $\mathrm{AnsF1}@1$ score of $48.4\%$ across four multi-hop benchmarks, outperforming all strong baselines, including the substantially larger ReSearch-32B ($46.2\%$). Extensive analyses further show that A$^2$Search resolves ambiguity and generalizes across benchmarks, highlighting that embracing ambiguity is essential for building more reliable QA systems. Our code, data, and model weights can be found at https://github.com/zfj1998/A2Search

---

### SimCast: Enhancing Precipitation Nowcasting with Short-to-Long Term Knowledge Distillation

Authors: Yifang Yin, Shengkai Chen, Yiyao Li, Lu Wang, Ruibing Jin, Wei Cui, Shili Xiang
Published: 2025-10-09
Categories: cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2510.07953v1

Precipitation nowcasting predicts future radar sequences based on current observations, which is a highly challenging task driven by the inherent complexity of the Earth system. Accurate nowcasting is of utmost importance for addressing various societal needs, including disaster management, agriculture, transportation, and energy optimization. As a complementary to existing non-autoregressive nowcasting approaches, we investigate the impact of prediction horizons on nowcasting models and propose SimCast, a novel training pipeline featuring a short-to-long term knowledge distillation technique coupled with a weighted MSE loss to prioritize heavy rainfall regions. Improved nowcasting predictions can be obtained without introducing additional overhead during inference. As SimCast generates deterministic predictions, we further integrate it into a diffusion-based framework named CasCast, leveraging the strengths from probabilistic models to overcome limitations such as blurriness and distribution shift in deterministic outputs. Extensive experimental results on three benchmark datasets validate the effectiveness of the proposed framework, achieving mean CSI scores of 0.452 on SEVIR, 0.474 on HKO-7, and 0.361 on MeteoNet, which outperforms existing approaches by a significant margin.

---

### A Large-scale Dataset for Robust Complex Anime Scene Text Detection

Authors: Ziyi Dong, Yurui Zhang, Changmao Li, Naomi Rue Golding, Qing Long
Published: 2025-10-09
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2510.07951v1

Current text detection datasets primarily target natural or document scenes, where text typically appear in regular font and shapes, monotonous colors, and orderly layouts. The text usually arranged along straight or curved lines. However, these characteristics differ significantly from anime scenes, where text is often diverse in style, irregularly arranged, and easily confused with complex visual elements such as symbols and decorative patterns. Text in anime scene also includes a large number of handwritten and stylized fonts. Motivated by this gap, we introduce AnimeText, a large-scale dataset containing 735K images and 4.2M annotated text blocks. It features hierarchical annotations and hard negative samples tailored for anime scenarios. %Cross-dataset evaluations using state-of-the-art methods demonstrate that models trained on AnimeText achieve superior performance in anime text detection tasks compared to existing datasets. To evaluate the robustness of AnimeText in complex anime scenes, we conducted cross-dataset benchmarking using state-of-the-art text detection methods. Experimental results demonstrate that models trained on AnimeText outperform those trained on existing datasets in anime scene text detection tasks. AnimeText on HuggingFace: https://huggingface.co/datasets/deepghs/AnimeText

---

### CVD-STORM: Cross-View Video Diffusion with Spatial-Temporal Reconstruction Model for Autonomous Driving

Authors: Tianrui Zhang, Yichen Liu, Zilin Guo, Yuxin Guo, Jingcheng Ni, Chenjing Ding, Dan Xu, Lewei Lu, Zehuan Wu
Published: 2025-10-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.07944v2

Generative models have been widely applied to world modeling for environment simulation and future state prediction. With advancements in autonomous driving, there is a growing demand not only for high-fidelity video generation under various controls, but also for producing diverse and meaningful information such as depth estimation. To address this, we propose CVD-STORM, a cross-view video diffusion model utilizing a spatial-temporal reconstruction Variational Autoencoder (VAE) that generates long-term, multi-view videos with 4D reconstruction capabilities under various control inputs. Our approach first fine-tunes the VAE with an auxiliary 4D reconstruction task, enhancing its ability to encode 3D structures and temporal dynamics. Subsequently, we integrate this VAE into the video diffusion process to significantly improve generation quality. Experimental results demonstrate that our model achieves substantial improvements in both FID and FVD metrics. Additionally, the jointly-trained Gaussian Splatting Decoder effectively reconstructs dynamic scenes, providing valuable geometric information for comprehensive scene understanding. Our project page is https://sensetime-fvg.github.io/CVD-STORM.

---

### Agent-Based Genetic Algorithm for Crypto Trading Strategy Optimization

Authors: Qiushi Tian, Churong Liang, Kairan Hong, Runnan Li
Published: 2025-10-09
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.07943v1

Cryptocurrency markets present formidable challenges for trading strategy optimization due to extreme volatility, non-stationary dynamics, and complex microstructure patterns that render conventional parameter optimization methods fundamentally inadequate. We introduce Cypto Genetic Algorithm Agent (CGA-Agent), a pioneering hybrid framework that synergistically integrates genetic algorithms with intelligent multi-agent coordination mechanisms for adaptive trading strategy parameter optimization in dynamic financial environments. The framework uniquely incorporates real-time market microstructure intelligence and adaptive strategy performance feedback through intelligent mechanisms that dynamically guide evolutionary processes, transcending the limitations of static optimization approaches. Comprehensive empirical evaluation across three cryptocurrencies demonstrates systematic and statistically significant performance improvements on both total returns and risk-adjusted metrics.

---

### TTOM: Test-Time Optimization and Memorization for Compositional Video Generation

Authors: Leigang Qu, Ziyang Wang, Na Zheng, Wenjie Wang, Liqiang Nie, Tat-Seng Chua
Published: 2025-10-09
Categories: cs.CV, cs.AI, cs.CL, cs.LG, cs.MM
arXiv: https://arxiv.org/abs/2510.07940v2

Video Foundation Models (VFMs) exhibit remarkable visual generation performance, but struggle in compositional scenarios (e.g., motion, numeracy, and spatial relation). In this work, we introduce Test-Time Optimization and Memorization (TTOM), a training-free framework that aligns VFM outputs with spatiotemporal layouts during inference for better text-image alignment. Rather than direct intervention to latents or attention per-sample in existing work, we integrate and optimize new parameters guided by a general layout-attention objective. Furthermore, we formulate video generation within a streaming setting, and maintain historical optimization contexts with a parametric memory mechanism that supports flexible operations, such as insert, read, update, and delete. Notably, we found that TTOM disentangles compositional world knowledge, showing powerful transferability and generalization. Experimental results on the T2V-CompBench and Vbench benchmarks establish TTOM as an effective, practical, scalable, and efficient framework to achieve cross-modal alignment for compositional video generation on the fly.

---

### Some theoretical improvements on the tightness of PAC-Bayes risk certificates for neural networks

Authors: Diego García-Pérez, Emilio Parrado-Hernández, John Shawe-Taylor
Published: 2025-10-09
Categories: cs.LG, cs.IT, stat.ML
arXiv: https://arxiv.org/abs/2510.07935v2

This paper presents four theoretical contributions that improve the usability of risk certificates for neural networks based on PAC-Bayes bounds. First, two bounds on the KL divergence between Bernoulli distributions enable the derivation of the tightest explicit bounds on the true risk of classifiers across different ranges of empirical risk. The paper next focuses on the formalization of an efficient methodology based on implicit differentiation that enables the introduction of the optimization of PAC-Bayesian risk certificates inside the loss/objective function used to fit the network/model. The last contribution is a method to optimize bounds on non-differentiable objectives such as the 0-1 loss. These theoretical contributions are complemented with an empirical evaluation on the MNIST and CIFAR-10 datasets. In fact, this paper presents the first non-vacuous generalization bounds on CIFAR-10 for neural networks. Code to reproduce all experiments is available at github.com/Diegogpcm/pacbayesgradients.

---

### Vision-Enabled LLMs in Historical Lexicography: Digitising and Enriching Estonian-German Dictionaries from the 17th and 18th Centuries

Authors: Madis Jürviste, Joonatan Jakobson
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.07931v1

This article presents research conducted at the Institute of the Estonian Language between 2022 and 2025 on the application of large language models (LLMs) to the study of 17th and 18th century Estonian dictionaries. The authors address three main areas: enriching historical dictionaries with modern word forms and meanings; using vision-enabled LLMs to perform text recognition on sources printed in Gothic script (Fraktur); and preparing for the creation of a unified, cross-source dataset. Initial experiments with J. Gutslaff's 1648 dictionary indicate that LLMs have significant potential for semi-automatic enrichment of dictionary information. When provided with sufficient context, Claude 3.7 Sonnet accurately provided meanings and modern equivalents for 81% of headword entries. In a text recognition experiment with A. T. Helle's 1732 dictionary, a zero-shot method successfully identified and structured 41% of headword entries into error-free JSON-formatted output. For digitising the Estonian-German dictionary section of A. W. Hupel's 1780 grammar, overlapping tiling of scanned image files is employed, with one LLM being used for text recognition and a second for merging the structured output. These findings demonstrate that even for minor languages LLMs have a significant potential for saving time and financial resources.

---

### ASBench: Image Anomalies Synthesis Benchmark for Anomaly Detection

Authors: Qunyi Zhang, Songan Zhang, Jiaqi Liu, Jinbao Wang, Xiaoning Lei, Guoyang Xie, Guannan Jiang, Zhichao Lu
Published: 2025-10-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.07927v3

Anomaly detection plays a pivotal role in manufacturing quality control, yet its application is constrained by limited abnormal samples and high manual annotation costs. While anomaly synthesis offers a promising solution, existing studies predominantly treat anomaly synthesis as an auxiliary component within anomaly detection frameworks, lacking systematic evaluation of anomaly synthesis algorithms. Current research also overlook crucial factors specific to anomaly synthesis, such as decoupling its impact from detection, quantitative analysis of synthetic data and adaptability across different scenarios. To address these limitations, we propose ASBench, the first comprehensive benchmarking framework dedicated to evaluating anomaly synthesis methods. Our framework introduces four critical evaluation dimensions: (i) the generalization performance across different datasets and pipelines (ii) the ratio of synthetic to real data (iii) the correlation between intrinsic metrics of synthesis images and anomaly detection performance metrics , and (iv) strategies for hybrid anomaly synthesis methods. Through extensive experiments, ASBench not only reveals limitations in current anomaly synthesis methods but also provides actionable insights for future research directions in anomaly synthesis

---

### Comprehensiveness Metrics for Automatic Evaluation of Factual Recall in Text Generation

Authors: Adam Dejl, James Barry, Alessandra Pascale, Javier Carnerero Cano
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.07926v2

Despite demonstrating remarkable performance across a wide range of tasks, large language models (LLMs) have also been found to frequently produce outputs that are incomplete or selectively omit key information. In sensitive domains, such omissions can result in significant harm comparable to that posed by factual inaccuracies, including hallucinations. In this study, we address the challenge of evaluating the comprehensiveness of LLM-generated texts, focusing on the detection of missing information or underrepresented viewpoints. We investigate three automated evaluation metrics: (1) an NLI-based method that decomposes texts into atomic statements and uses natural language inference (NLI) to identify missing facts, (2) a Q&A-based metric that extracts question-answer pairs and compares responses across sources, and (3) an end-to-end approach that directly identifies missing content using LLMs. Our experiments demonstrate the surprising effectiveness of the simple end-to-end metric compared to more complex metrics, though at the cost of reduced robustness, interpretability and result granularity. We further assess the comprehensiveness of responses from several popular open-weight LLMs when answering user queries based on multiple sources.

---

### Enabling Personalized Long-term Interactions in LLM-based Agents through Persistent Memory and User Profiles

Authors: Rebecca Westhäußer, Wolfgang Minker, Sebatian Zepf
Published: 2025-10-09
Categories: cs.AI, cs.HC
arXiv: https://arxiv.org/abs/2510.07925v1

Large language models (LLMs) increasingly serve as the central control unit of AI agents, yet current approaches remain limited in their ability to deliver personalized interactions. While Retrieval Augmented Generation enhances LLM capabilities by improving context-awareness, it lacks mechanisms to combine contextual information with user-specific data. Although personalization has been studied in fields such as human-computer interaction or cognitive science, existing perspectives largely remain conceptual, with limited focus on technical implementation. To address these gaps, we build on a unified definition of personalization as a conceptual foundation to derive technical requirements for adaptive, user-centered LLM-based agents. Combined with established agentic AI patterns such as multi-agent collaboration or multi-source retrieval, we present a framework that integrates persistent memory, dynamic coordination, self-validation, and evolving user profiles to enable personalized long-term interactions. We evaluate our approach on three public datasets using metrics such as retrieval accuracy, response correctness, or BertScore. We complement these results with a five-day pilot user study providing initial insights into user feedback on perceived personalization. The study provides early indications that guide future work and highlights the potential of integrating persistent memory and user profiles to improve the adaptivity and perceived personalization of LLM-based agents.

---

### Synergy Between the Strong and the Weak: Spiking Neural Networks are Inherently Self-Distillers

Authors: Yongqi Ding, Lin Zuo, Mengmeng Jing, Kunshan Yang, Pei He, Tonglan Xie
Published: 2025-10-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.07924v1

Brain-inspired spiking neural networks (SNNs) promise to be a low-power alternative to computationally intensive artificial neural networks (ANNs), although performance gaps persist. Recent studies have improved the performance of SNNs through knowledge distillation, but rely on large teacher models or introduce additional training overhead. In this paper, we show that SNNs can be naturally deconstructed into multiple submodels for efficient self-distillation. We treat each timestep instance of the SNN as a submodel and evaluate its output confidence, thus efficiently identifying the strong and the weak. Based on this strong and weak relationship, we propose two efficient self-distillation schemes: (1) \textbf{Strong2Weak}: During training, the stronger "teacher" guides the weaker "student", effectively improving overall performance. (2) \textbf{Weak2Strong}: The weak serve as the "teacher", distilling the strong in reverse with underlying dark knowledge, again yielding significant performance gains. For both distillation schemes, we offer flexible implementations such as ensemble, simultaneous, and cascade distillation. Experiments show that our method effectively improves the discriminability and overall performance of the SNN, while its adversarial robustness is also enhanced, benefiting from the stability brought by self-distillation. This ingeniously exploits the temporal properties of SNNs and provides insight into how to efficiently train high-performance SNNs.

---

### STEPER: Step-wise Knowledge Distillation for Enhancing Reasoning Ability in Multi-Step Retrieval-Augmented Language Models

Authors: Kyumin Lee, Minjin Jeon, Sanghwan Jang, Hwanjo Yu
Published: 2025-10-09
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.07923v1

Answering complex real-world questions requires step-by-step retrieval and integration of relevant information to generate well-grounded responses. However, existing knowledge distillation methods overlook the need for different reasoning abilities at different steps, hindering transfer in multi-step retrieval-augmented frameworks. To address this, we propose Stepwise Knowledge Distillation for Enhancing Reasoning Ability in Multi-Step Retrieval-Augmented Language Models (StepER). StepER employs step-wise supervision to align with evolving information and reasoning demands across stages. Additionally, it incorporates difficulty-aware training to progressively optimize learning by prioritizing suitable steps. Our method is adaptable to various multi-step retrieval-augmented language models, including those that use retrieval queries for reasoning paths or decomposed questions. Extensive experiments show that StepER outperforms prior methods on multi-hop QA benchmarks, with an 8B model achieving performance comparable to a 70B teacher model.

---

### Screen Before You Fetch: Compressed Byzantine Screening for Decentralized Learning on the Edge-Cloud Continuum

Authors: Murtaza Rangwala, Farag Azzedin, Richard O. Sinnott, Rajkumar Buyya
Published: 2025-10-09
Categories: cs.LG, cs.DC
arXiv: https://arxiv.org/abs/2510.07922v6

In decentralized federated learning, Byzantine-robust defenses filter neighbors by model similarity, but only after receiving each neighbor's full model. On constrained edge uplinks, that transfer dominates the round, even for discarded models. We propose SketchGuard, which screens neighbors on compact sketches and fetches full models only from those that pass, so a rejected neighbor costs $O(k)$ rather than $O(d)$ where $k\ll d$. Sketches, however, hide information an adversary can exploit, either by building a model indistinguishable from a benign one under a public compressor or by forging a sketch once the seed is known. Three rules close these gaps, namely exact verification of every fetched model, commitment before the sketch seed is drawn, and a ban on any sender whose model mismatches its sketch. We prove convergence in the strongly convex and non-convex settings with explicit topology and heterogeneity terms. Across three datasets and two wrapped filters, SketchGuard reproduces the uncompressed filter's test error to within $0.003$, withstands a band-targeted attack that defeats sketch-only screening, and moves $0.7\times$ the bytes of full exchange under attack for under $1.7\%$ of a round in compute. On emulated edge and cloud links it cuts the round to $0.61\times$ that of the state-of-the-art.

---

### Profit Mirage: Revisiting Information Leakage in LLM-based Financial Agents

Authors: Xiangyu Li, Yawen Zeng, Xiaofen Xing, Jin Xu, Xiangmin Xu
Published: 2025-10-09
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.07920v1

LLM-based financial agents have attracted widespread excitement for their ability to trade like human experts. However, most systems exhibit a "profit mirage": dazzling back-tested returns evaporate once the model's knowledge window ends, because of the inherent information leakage in LLMs. In this paper, we systematically quantify this leakage issue across four dimensions and release FinLake-Bench, a leakage-robust evaluation benchmark. Furthermore, to mitigate this issue, we introduce FactFin, a framework that applies counterfactual perturbations to compel LLM-based agents to learn causal drivers instead of memorized outcomes. FactFin integrates four core components: Strategy Code Generator, Retrieval-Augmented Generation, Monte Carlo Tree Search, and Counterfactual Simulator. Extensive experiments show that our method surpasses all baselines in out-of-sample generalization, delivering superior risk-adjusted performance.

---

### GRADE: Personalized Multi-Task Fusion via Group-relative Reinforcement Learning with Adaptive Dirichlet Exploration

Authors: Tingfeng Hong, Pingye Ren, Xinlong Xiao, Chao Wang, Chenyi Lei, Wenwu Ou, Han Li
Published: 2025-10-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.07919v2

Balancing multiple objectives is critical for user satisfaction in modern recommender and search systems, yet current Multi-Task Fusion (MTF) methods rely on static, manually-tuned weights that fail to capture individual user intent. While Reinforcement Learning (RL) offers a path to personalization, traditional approaches often falter due to training instability and the sparse rewards inherent in these large-scale systems. To address these limitations, we propose Group-relative Reinforcement learning with Adaptive Dirichlet Exploration (GRADE), a novel and robust framework for personalized multi-task fusion. GRADE leverages a critic-free, Group Relative Policy Optimization (GRPO) paradigm, enabling stable and efficient policy learning by evaluating the relative performance of candidate weight groups. Its core innovations include employing the Dirichlet distribution for principled and structured exploration of the weight space, and a composite reward function that combines sparse user feedback with dense model priors and rule-based constraints to guide the search effectively. Deployed in the in-app marketplace of an application with over hundreds of millions daily active users, GRADE significantly outperforms established baselines, achieving substantial gains in rigorous large-scale A/B tests: +0.595\% in CTR, +1.193\% in CVR, +1.788\% in OPM, and +1.568\% in total order volume. Following its strong performance, GRADE has been fully deployed in the marketplace search scenario of Kuaishou, serving hundreds of millions of users.

---

### MARC: Memory-Augmented RL Token Compression for Efficient Video Understanding

Authors: Peiran Wu, Zhuorui Yu, Yunze Liu, Chi-Hao Wu, Enmin Zhou, Junxiao Shen
Published: 2025-10-09
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.07915v3

The rapid progress of large language models (LLMs) has laid the foundation for multimodal models. However, visual language models (VLMs) still face heavy computational costs when extended from images to videos due to high frame rates and long durations. Token compression is a promising solution, yet most existing training-free methods cause information loss and performance degradation. To overcome this, we propose \textbf{Memory-Augmented Reinforcement Learning-based Token Compression (MARC)}, which integrates structured retrieval and RL-based distillation. MARC adopts a \textit{retrieve-then-compress} strategy using a \textbf{Visual Memory Retriever (VMR)} to select key clips and a \textbf{Compression Group Relative Policy Optimization (C-GRPO)} framework to distil reasoning ability from a teacher to a student model. Experiments on six video benchmarks show that MARC achieves near-baseline accuracy using only one frame's tokens -- reducing visual tokens by \textbf{95\%}, GPU memory by \textbf{72\%}, and latency by \textbf{23.9\%}. This demonstrates its potential for efficient, real-time video understanding in resource-constrained settings such as video QA, surveillance, and autonomous driving.

---

### Towards Human-Like Grading: A Unified LLM-Enhanced Framework for Subjective Question Evaluation

Authors: Fanwei Zhua, Jiaxuan He, Xiaoxiao Chen, Zulong Chen, Quan Lu, Chenrui Mei
Published: 2025-10-09
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.07912v1

Automatic grading of subjective questions remains a significant challenge in examination assessment due to the diversity in question formats and the open-ended nature of student responses. Existing works primarily focus on a specific type of subjective question and lack the generality to support comprehensive exams that contain diverse question types. In this paper, we propose a unified Large Language Model (LLM)-enhanced auto-grading framework that provides human-like evaluation for all types of subjective questions across various domains. Our framework integrates four complementary modules to holistically evaluate student answers. In addition to a basic text matching module that provides a foundational assessment of content similarity, we leverage the powerful reasoning and generative capabilities of LLMs to: (1) compare key knowledge points extracted from both student and reference answers, (2) generate a pseudo-question from the student answer to assess its relevance to the original question, and (3) simulate human evaluation by identifying content-related and non-content strengths and weaknesses. Extensive experiments on both general-purpose and domain-specific datasets show that our framework consistently outperforms traditional and LLM-based baselines across multiple grading metrics. Moreover, the proposed system has been successfully deployed in real-world training and certification exams at a major e-commerce enterprise.

---

### MMM: Quantum-Chemical Molecular Representation Learning for Combinatorial Drug Recommendation

Authors: Chongmyung Kwon, Yujin Kim, Seoeun Park, Yunji Lee, Charmgil Hong
Published: 2025-10-09
Categories: cs.LG, cs.AI, cs.CV
arXiv: https://arxiv.org/abs/2510.07910v1

Drug recommendation is an essential task in machine learning-based clinical decision support systems. However, the risk of drug-drug interactions (DDI) between co-prescribed medications remains a significant challenge. Previous studies have used graph neural networks (GNNs) to represent drug structures. Regardless, their simplified discrete forms cannot fully capture the molecular binding affinity and reactivity. Therefore, we propose Multimodal DDI Prediction with Molecular Electron Localization Function (ELF) Maps (MMM), a novel framework that integrates three-dimensional (3D) quantum-chemical information into drug representation learning. It generates 3D electron density maps using the ELF. To capture both therapeutic relevance and interaction risks, MMM combines ELF-derived features that encode global electronic properties with a bipartite graph encoder that models local substructure interactions. This design enables learning complementary characteristics of drug molecules. We evaluate MMM in the MIMIC-III dataset (250 drugs, 442 substructures), comparing it with several baseline models. In particular, a comparison with the GNN-based SafeDrug model demonstrates statistically significant improvements in the F1-score (p = 0.0387), Jaccard (p = 0.0112), and the DDI rate (p = 0.0386). These results demonstrate the potential of ELF-based 3D representations to enhance prediction accuracy and support safer combinatorial drug prescribing in clinical practice.

---

### SatFusion: A Unified Framework for Enhancing Remote Sensing Images via Multi-Frame and Multi-Source Images Fusion

Authors: Yufei Tong, Guanjie Cheng, Peihan Wu, Feiyi Chen, Xinkui Zhao, Shuiguang Deng
Published: 2025-10-09
Categories: eess.IV, cs.CV, cs.MM
arXiv: https://arxiv.org/abs/2510.07905v4

High-quality remote sensing (RS) image acquisition is fundamentally constrained by physical limitations. While Multi-Frame Super-Resolution (MFSR) and Pansharpening address this by exploiting complementary information, they are typically studied in isolation: MFSR lacks high-resolution (HR) structural priors for fine-grained texture recovery, whereas Pansharpening relies on upsampled low-resolution (LR) inputs and is sensitive to noise and misalignment. In this paper, we propose SatFusion, a novel and unified framework that seamlessly bridges multi-frame and multi-source RS image fusion. SatFusion extracts HR semantic features by aggregating complementary information from multiple LR multispectral frames via a Multi-Frame Image Fusion (MFIF) module, and integrates fine-grained structural details from an HR panchromatic image through a Multi-Source Image Fusion (MSIF) module with implicit pixel-level alignment. To further alleviate the lack of structural priors during multi-frame fusion, we introduce an advanced variant, SatFusion*, which integrates a panchromatic-guided mechanism into the MFIF stage. Through structure-aware feature embedding and transformer-based adaptive aggregation, SatFusion* enables spatially adaptive feature selection, strengthening the coupling between multi-frame and multi-source representations. Extensive experiments on four benchmark datasets validate our core insight: synergistically coupling multi-frame and multi-source priors effectively resolves the fragility of existing paradigms, delivering superior reconstruction fidelity, robustness, and generalizability.

---

### Multi-level informed optimization via decomposed Kriging for large design problems under uncertainty

Authors: Enrico Ampellio, Blazhe Gjorgiev, Giovanni Sansavini
Published: 2025-10-09
Categories: eess.SY, cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2510.07904v1

Engineering design involves demanding models encompassing many decision variables and uncontrollable parameters. In addition, unavoidable aleatoric and epistemic uncertainties can be very impactful and add further complexity. The state-of-the-art adopts two steps, uncertainty quantification and design optimization, to optimize systems under uncertainty by means of robust or stochastic metrics. However, conventional scenario-based, surrogate-assisted, and mathematical programming methods are not sufficiently scalable to be affordable and precise in large and complex cases. Here, a multi-level approach is proposed to accurately optimize resource-intensive, high-dimensional, and complex engineering problems under uncertainty with minimal resources. A non-intrusive, fast-scaling, Kriging-based surrogate is developed to map the combined design/parameter domain efficiently. Multiple surrogates are adaptively updated by hierarchical and orthogonal decomposition to leverage the fewer and most uncertainty-informed data. The proposed method is statistically compared to the state-of-the-art via an analytical testbed and is shown to be concurrently faster and more accurate by orders of magnitude.

---

### ACE: Attribution-Controlled Knowledge Editing for Multi-hop Factual Recall

Authors: Jiayu Yang, Yuxuan Fan, Songning Lai, Shengen Wu, Jiaqi Tang, Chun Kang, Zhijiang Guo, Yutao Yue
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.07896v3

Large Language Models (LLMs) require efficient knowledge editing (KE) to update factual information, yet existing methods exhibit significant performance decay in multi-hop factual recall. This failure is particularly acute when edits involve intermediate implicit subjects within reasoning chains. Through causal analysis, we reveal that this limitation stems from an oversight of how chained knowledge is dynamically represented and utilized at the neuron level. We discover that during multi hop reasoning, implicit subjects function as query neurons, which sequentially activate corresponding value neurons across transformer layers to accumulate information toward the final answer, a dynamic prior KE work has overlooked. Guided by this insight, we propose ACE: Attribution-Controlled Knowledge Editing for Multi-hop Factual Recall, a framework that leverages neuron-level attribution to identify and edit these critical query-value (Q-V) pathways. ACE provides a mechanistically grounded solution for multi-hop KE, empirically outperforming state-of-the-art methods by 9.44% on GPT-J and 37.46% on Qwen3-8B. Our analysis further reveals more fine-grained activation patterns in Qwen3 and demonstrates that the semantic interpretability of value neurons is orchestrated by query-driven accumulation. These findings establish a new pathway for advancing KE capabilities based on the principled understanding of internal reasoning mechanisms.

---

### Adaptive Optimizable Gaussian Process Regression Linear Least Squares Regression Filtering Method for SEM Images

Authors: D. Chee Yong Ong, I. Bukhori, K. S. Sim, K. Beng Gan
Published: 2025-10-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.07895v1

Scanning Electron Microscopy (SEM) images often suffer from noise contamination, which degrades image quality and affects further analysis. This research presents a complete approach to estimate their Signal-to-Noise Ratio (SNR) and noise variance (NV), and enhance image quality using NV-guided Wiener filter. The main idea of this study is to use a good SNR estimation technique and infuse a machine learning model to estimate NV of the SEM image, which then guides the wiener filter to remove the noise, providing a more robust and accurate SEM image filtering pipeline. First, we investigate five different SNR estimation techniques, namely Nearest Neighbourhood (NN) method, First-Order Linear Interpolation (FOL) method, Nearest Neighbourhood with First-Order Linear Interpolation (NN+FOL) method, Non-Linear Least Squares Regression (NLLSR) method, and Linear Least Squares Regression (LSR) method. It is shown that LSR method to perform better than the rest. Then, Support Vector Machines (SVM) and Gaussian Process Regression (GPR) are tested by pairing it with LSR. In this test, the Optimizable GPR model shows the highest accuracy and it stands as the most effective solution for NV estimation. Combining these results lead to the proposed Adaptive Optimizable Gaussian Process Regression Linear Least Squares Regression (AO-GPRLLSR) Filtering pipeline. The AO-GPRLLSR method generated an estimated noise variance which served as input to NV-guided Wiener filter for improving the quality of SEM images. The proposed method is shown to achieve notable success in estimating SNR and NV of SEM images and leads to lower Mean Squared Error (MSE) after the filtering process.

---

### Metric Calculating Benchmark: Code-Verifiable Complicate Instruction Following Benchmark for Large Language Models

Authors: Hyeonseok Moon, Seongtae Hong, Jaehyung Seo, Heuiseok Lim
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.07892v1

Recent frontier-level LLMs have saturated many previously difficult benchmarks, leaving little room for further differentiation. This progress highlights the need for challenging benchmarks that provide objective verification. In this paper, we introduce MCBench, a benchmark designed to evaluate whether LLMs can execute string-matching NLP metrics by strictly following step-by-step instructions. Unlike prior benchmarks that depend on subjective judgments or general reasoning, MCBench offers an objective, deterministic and codeverifiable evaluation. This setup allows us to systematically test whether LLMs can maintain accurate step-by-step execution, including instruction adherence, numerical computation, and long-range consistency in handling intermediate results. To ensure objective evaluation of these abilities, we provide a parallel reference code that can evaluate the accuracy of LLM output. We provide three evaluative metrics and three benchmark variants designed to measure the detailed instruction understanding capability of LLMs. Our analyses show that MCBench serves as an effective and objective tool for evaluating the capabilities of cutting-edge LLMs.

---

### Standard-to-Dialect Transfer Trends Differ across Text and Speech: A Case Study on Intent and Topic Classification in German Dialects

Authors: Verena Blaschke, Miriam Winkler, Barbara Plank
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.07890v3

Research on cross-dialectal transfer from a standard to a non-standard dialect variety has typically focused on text data. However, dialects are primarily spoken, and non-standard spellings cause issues in text processing. We compare standard-to-dialect transfer in three settings: text models, speech models, and cascaded systems where speech first gets automatically transcribed and then further processed by a text model. We focus on German dialects in the context of written and spoken intent classification -- releasing the first dialectal audio intent classification dataset -- with supporting experiments on topic classification. The speech-only setup provides the best results on the dialect data while the text-only setup works best on the standard data. While the cascaded systems lag behind the text-only models for German, they perform relatively well on the dialectal data if the transcription system generates normalized, standard-like output.

---

### Towards Meaningful Transparency in Civic AI Systems

Authors: Dave Murray-Rust, Kars Alfrink, Cristina Zaga
Published: 2025-10-09
Categories: cs.AI, cs.CY, cs.HC
arXiv: https://arxiv.org/abs/2510.07889v1

Artificial intelligence has become a part of the provision of governmental services, from making decisions about benefits to issuing fines for parking violations. However, AI systems rarely live up to the promise of neutral optimisation, creating biased or incorrect outputs and reducing the agency of both citizens and civic workers to shape the way decisions are made. Transparency is a principle that can both help subjects understand decisions made about them and shape the processes behind those decisions. However, transparency as practiced around AI systems tends to focus on the production of technical objects that represent algorithmic aspects of decision making. These are often difficult for publics to understand, do not connect to potential for action, and do not give insight into the wider socio-material context of decision making. In this paper, we build on existing approaches that take a human-centric view on AI transparency, combined with a socio-technical systems view, to develop the concept of meaningful transparency for civic AI systems: transparencies that allow publics to engage with AI systems that affect their lives, connecting understanding with potential for action.

---

### Signal-to-Noise Ratio in Scanning Electron Microscopy: A Comprehensive Review

Authors: K. S. Sim, I. Bukhori, D. C. Y. Ong, K. B. Gan
Published: 2025-10-09
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.07886v1

Scanning Electron Microscopy (SEM) is critical in nanotechnology, materials science, and biological imaging due to its high spatial resolution and depth of focus. Signal-to-noise ratio (SNR) is an essential parameter in SEM because it directly impacts the quality and interpretability of the images. SEM is widely used in various scientific disciplines, but its utility can be compromised by noise, which degrades image clarity. This review explores multiple aspects of the SEM imaging process, from the principal operation of SEM, sources of noise in SEM, methods for SNR measurement and estimations, to various aspects that affect the SNR measurement and approaches to enhance SNR, both from a hardware and software standpoint. We review traditional and emerging techniques, focusing on their applications, advantages, and limitations. The paper aims to provide a comprehensive understanding of SNR optimization in SEM for researchers and practitioners and to encourage further research in the field.

---

### Contrastive Weak-to-strong Generalization

Authors: Houcheng Jiang, Junfeng Fang, Jiaxin Wu, Tianyu Zhang, Chen Gao, Xiang Wang, Xiangnan He, Yang Deng
Published: 2025-10-09
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.07884v3

Weak-to-strong generalization provides a promising paradigm for scaling large language models (LLMs) by training stronger models on samples from aligned weaker ones, without requiring human feedback or explicit reward modeling. However, its robustness and generalization are hindered by the noise and biases in weak-model outputs, which limit its applicability in practice. To address this challenge, we leverage implicit rewards, which approximate explicit rewards through log-likelihood ratios, and reveal their structural equivalence with Contrastive Decoding (CD), a decoding strategy shown to reduce noise in LLM generation. Building on this connection, we propose Contrastive Weak-to-Strong Generalization (ConG), a framework that employs contrastive decoding between pre- and post-alignment weak models to generate higher-quality samples. This approach enables more reliable capability transfer, denoising, and improved robustness, substantially mitigating the limitations of traditional weak-to-strong methods. Empirical results across different model families confirm consistent improvements, demonstrating the generality and effectiveness of ConG. Taken together, our findings highlight the potential of ConG to advance weak-to-strong generalization and provide a promising pathway toward AGI.

---

### CS3-Bench: Evaluating and Enhancing Speech-to-Speech LLMs for Mandarin-English Code-Switching

Authors: Heyang Liu, Yuhao Wang, Ziyang Cheng, Ronghua Wu, Qunshan Gu, Yanfeng Wang, Yu Wang
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.07881v1

The advancement of multimodal large language models has accelerated the development of speech-to-speech interaction systems. While natural monolingual interaction has been achieved, we find existing models exhibit deficiencies in language alignment. In our proposed Code-Switching Speech-to-Speech Benchmark (CS3-Bench), experiments on 7 mainstream models demonstrate a relative performance drop of up to 66% in knowledge-intensive question answering and varying degrees of misunderstanding in open-ended conversations. Starting from a model with severe performance deterioration, we propose both data constructions and training approaches to improve the language alignment capabilities, specifically employing Chain of Recognition (CoR) to enhance understanding and Keyword Highlighting (KH) to guide generation. Our approach improves the knowledge accuracy from 25.14% to 46.13%, with open-ended understanding rate from 64.5% to 86.5%, and significantly reduces pronunciation errors in the secondary language. CS3-Bench is available at https://huggingface.co/datasets/VocalNet/CS3-Bench.

---

### Do LLMs Really Need 10+ Thoughts for "Find the Time 1000 Days Later"? Towards Structural Understanding of LLM Overthinking

Authors: Xinliang Frederick Zhang, Anhad Mohananey, Alexandra Chronopoulou, Pinelopi Papalampidi, Somit Gupta, Tsendsuren Munkhdalai, Lu Wang, Shyam Upadhyay
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.07880v2

Models employing long chain-of-thought (CoT) reasoning have shown superior performance on complex reasoning tasks. Yet, this capability introduces a critical and often overlooked inefficiency -- overthinking -- models often engage in unnecessarily extensive reasoning even for simple queries, incurring significant computations without accuracy improvements. While prior work has explored solutions to mitigate overthinking, a fundamental gap remains in our understanding of its underlying causes. Most existing analyses are limited to superficial, profiling-based observations, failing to delve into LLMs' inner workings. This study introduces a systematic, fine-grained analyzer of LLMs' thought process to bridge the gap, TRACE. We first benchmark the overthinking issue, confirming that long-thinking models are five to twenty times slower on simple tasks with no substantial gains. We then use TRACE to first decompose the thought process into minimally complete sub-thoughts. Next, by inferring discourse relationships among sub-thoughts, we construct granular thought progression graphs and subsequently identify common thinking patterns for topically similar queries. Our analysis reveals two major patterns for open-weight thinking models -- Explorer and Late Landing. This finding provides evidence that over-verification and over-exploration are the primary drivers of overthinking in LLMs. Grounded in thought structures, we propose a utility-based definition of overthinking, which moves beyond length-based metrics. This revised definition offers a more insightful understanding of LLMs' thought progression, as well as practical guidelines for principled overthinking management.

---

### FlowLensing: Simulating Gravitational Lensing with Flow Matching

Authors: Hamees Sayed, Pranath Reddy, Michael W. Toomey, Sergei Gleyzer
Published: 2025-10-09
Categories: astro-ph.IM, cs.CV
arXiv: https://arxiv.org/abs/2510.07878v3

Gravitational lensing is one of the most powerful probes of dark matter, yet creating high-fidelity lensed images at scale remains a bottleneck. Existing tools rely on ray-tracing or forward-modeling pipelines that, while precise, are prohibitively slow. We introduce FlowLensing, a Diffusion Transformer-based compact and efficient flow-matching model for strong gravitational lensing simulation. FlowLensing operates in both discrete and continuous regimes, handling classes such as different dark matter models as well as continuous model parameters ensuring physical consistency. By enabling scalable simulations, our model can advance dark matter studies, specifically for probing dark matter substructure in cosmological surveys. We find that our model achieves a speedup of over 200$\times$ compared to classical simulators for intensive dark matter models, with high fidelity and low inference latency. FlowLensing enables rapid, scalable, and physically consistent image synthesis, offering a practical alternative to traditional forward-modeling pipelines.

---

### Ready to Translate, Not to Represent? Bias and Performance Gaps in Multilingual LLMs Across Language Families and Domains

Authors: Md. Faiyaz Abdullah Sayeedi, Md. Mahbub Alam, Subhey Sadi Rahman, Md. Adnanul Islam, Jannatul Ferdous Deepti, Tasnim Mohiuddin, Md Mofijul Islam, Swakkhar Shatabda
Published: 2025-10-09
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.07877v2

The rise of Large Language Models (LLMs) has redefined Machine Translation (MT), enabling context-aware and fluent translations across hundreds of languages and textual domains. Despite their remarkable capabilities, LLMs often exhibit uneven performance across language families and specialized domains. Moreover, recent evidence reveals that these models can encode and amplify different biases present in their training data, posing serious concerns for fairness, especially in low-resource languages. To address these gaps, we introduce Translation Tangles, a unified framework and dataset for evaluating the translation quality and fairness of open-source LLMs. Our approach benchmarks 24 bidirectional language pairs across multiple domains using different metrics. We further propose a hybrid bias detection pipeline that integrates rule-based heuristics, semantic similarity filtering, and LLM-based validation. We also introduce a high-quality, bias-annotated dataset based on human evaluations of 1,439 translation-reference pairs. The code and dataset are accessible on GitHub: https://github.com/faiyazabdullah/TranslationTangles

---

### Learning to Navigate Socially Through Proactive Risk Perception

Authors: Erjia Xiao, Lingfeng Zhang, Yingbo Tang, Hao Cheng, Renjing Xu, Wenbo Ding, Lei Zhou, Long Chen, Hangjun Ye, Xiaoshuai Hao
Published: 2025-10-09
Categories: cs.RO, cs.AI, cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2510.07871v3

In this report, we describe the technical details of our submission to the IROS 2025 RoboSense Challenge Social Navigation Track. This track focuses on developing RGBD-based perception and navigation systems that enable autonomous agents to navigate safely, efficiently, and socially compliantly in dynamic human-populated indoor environments. The challenge requires agents to operate from an egocentric perspective using only onboard sensors including RGB-D observations and odometry, without access to global maps or privileged information, while maintaining social norm compliance such as safe distances and collision avoidance. Building upon the Falcon model, we introduce a Proactive Risk Perception Module to enhance social navigation performance. Our approach augments Falcon with collision risk understanding that learns to predict distance-based collision risk scores for surrounding humans, which enables the agent to develop more robust spatial awareness and proactive collision avoidance behaviors. The evaluation on the Social-HM3D benchmark demonstrates that our method improves the agent's ability to maintain personal space compliance while navigating toward goals in crowded indoor scenes with dynamic human agents, achieving 2nd place among 16 participating teams in the challenge.

---

### On the Optimality of the Median-of-Means Estimator under Adversarial Contamination

Authors: Xabier de Juan, Santiago Mazuelas
Published: 2025-10-09
Categories: stat.ML, cs.LG, math.ST
arXiv: https://arxiv.org/abs/2510.07867v1

The Median-of-Means (MoM) is a robust estimator widely used in machine learning that is known to be (minimax) optimal in scenarios where samples are i.i.d. In more grave scenarios, samples are contaminated by an adversary that can inspect and modify the data. Previous work has theoretically shown the suitability of the MoM estimator in certain contaminated settings. However, the (minimax) optimality of MoM and its limitations under adversarial contamination remain unknown beyond the Gaussian case. In this paper, we present upper and lower bounds for the error of MoM under adversarial contamination for multiple classes of distributions. In particular, we show that MoM is (minimax) optimal in the class of distributions with finite variance, as well as in the class of distributions with infinite variance and finite absolute $(1+r)$-th moment. We also provide lower bounds for MoM's error that match the order of the presented upper bounds, and show that MoM is sub-optimal for light-tailed distributions.

---