# Research Papers — 2025-10 · bundle page 213 of 217

> Full-text markdown of papers #10601–10650 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/213
[← Previous bundle](https://aipod.dev/papers/archive/2025-10/page/212.md)
[Next bundle →](https://aipod.dev/papers/archive/2025-10/page/214.md)

### Resolution as a Direction: Vector-Panning Feature Alignment for Cross-Resolution Re-Identification

Authors: Zanwu Liu, Chao Yuan, Bo Li, Xiaowei Zhang, Guanglin Niu
Published: 2025-10-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.00936v2

Cross-resolution person re-identification (CR-ReID) remains challenging in practical surveillance, where camera quality and capture distance lead to substantial resolution gaps between low-resolution (LR) queries and high-resolution (HR) gallery images. Prior approaches commonly rely on super-resolution (SR) or resolution-invariant representation learning, which often increases system complexity and may not directly address the feature mismatch induced by resolution degradation. In this work, we report a new empirical finding from a dedicated analysis in which identity-specific variation is averaged out: the HR--LR feature discrepancy produced by standard ReID backbones exhibits a consistent, resolution-related semantic direction in the embedding space. We further support this observation with statistical analyses based on Canonical Correlation Analysis (CCA) and Pearson correlation analysis. Motivated by this finding, we propose Vector Panning Feature Alignment (VPFA), a lightweight post-hoc module that learns to pan LR features along the learned resolution direction to obtain pseudo-HR representations. VPFA operates after feature extraction and can be integrated into existing ReID systems with negligible overhead. Extensive experiments on multiple CR-ReID benchmarks show that VPFA achieves state-of-the-art performance while improving efficiency compared to SR-based or jointly trained alternatives.

---

### Making, not Taking, the Best of N

Authors: Ammar Khairi, Daniel D'souza, Marzieh Fadaee, Julia Kreutzer
Published: 2025-10-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.00931v1

Obtaining high-quality generations in modern LLMs has largely been framed as a selection problem: identifying a single winning generation from a diverse pool of N samples, the Best-of-N (BoN). Yet, this approach is inherently zero-sum, discarding diverse and potentially useful information from the pool. Instead, we explore a collaborative setup, where all candidates can potentially contribute to the final winning generation. To this end, we propose Fusion-of-N (FusioN): a method that uses a general LLM judge to synthesize the most informative elements of each sample into a single final answer. We compare FusioN to BoN in two settings, (i) test-time scaling, where we sample and aggregate from a single model at test-time (ii) synthetic data generation, where we fuse samples from a pool of diverse teachers to improve a student model. We extensively benchmark both setups across 11 languages, 3 diverse tasks and varying model scales. Across the bench, FusioN consistently outperforms BoN showing versatility and robustness both in test-time scaling and in downstream gains from synthetic data generation. We also perform extensive analysis on FusioN, where it shows surprising strengths and robustness under challenging settings. These results show that we should shift how we think about evaluating and utilizing LLM generations from a monolithic measure of quality, to embracing their polylithic nature. This shift allows us to integrate diverse strengths, unlock latent potential, and achieve improvements that were previously inaccessible through selection alone.

---

### Equivariant Splitting: Self-supervised learning from incomplete data

Authors: Victor Sechaud, Jérémy Scanvic, Quentin Barthélemy, Patrice Abry, Julián Tachella
Published: 2025-10-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.00929v6

Self-supervised learning for inverse problems allows to train a reconstruction network from noise and/or incomplete data alone. These methods have the potential of enabling learning-based solutions when obtaining ground-truth references for training is expensive or even impossible. In this paper, we propose a new self-supervised learning strategy devised for the challenging setting where measurements are observed via a single incomplete observation model. We introduce a new definition of equivariance in the context of reconstruction networks, and show that the combination of self-supervised splitting losses and equivariant reconstruction networks results in unbiased estimates of the supervised loss. Through a series of experiments on image inpainting, accelerated magnetic resonance imaging, sparse-view computed tomography, and compressive sensing, we demonstrate that the proposed loss achieves state-of-the-art performance in settings with highly rank-deficient forward models. The code is available at https://github.com/vsechaud/Equivariant-Splitting

---

### On Discovering Algorithms for Adversarial Imitation Learning

Authors: Shashank Reddy Chirra, Jayden Teoh, Praveen Paruchuri, Pradeep Varakantham
Published: 2025-10-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.00922v2

Adversarial Imitation Learning (AIL) methods, while effective in settings with limited expert demonstrations, are often considered unstable. These approaches typically decompose into two components: Density Ratio (DR) estimation $\frac{ρ_E}{ρ_π}$, where a discriminator estimates the relative occupancy of state-action pairs under the policy versus the expert; and Reward Assignment (RA), where this ratio is transformed into a reward signal used to train the policy. While significant research has focused on improving density estimation, the role of reward assignment in influencing training dynamics and final policy performance has been largely overlooked. RA functions in AIL are typically derived from divergence minimization objectives, relying heavily on human design and ingenuity. In this work, we take a different approach: we investigate the discovery of data-driven RA functions, i.e, based directly on the performance of the resulting imitation policy. To this end, we leverage an LLM-guided evolutionary framework that efficiently explores the space of RA functions, yielding \emph{Discovered Adversarial Imitation Learning} (DAIL), the first meta-learnt AIL algorithm. Remarkably, DAIL generalises across unseen environments and policy optimization algorithms, outperforming the current state-of-the-art of \emph{human-designed} baselines. Finally, we analyse why DAIL leads to more stable training, offering novel insights into the role of RA functions in the stability of AIL. Code is publicly available: https://github.com/shshnkreddy/DAIL.

---

### Benchmarking Foundation Models with Retrieval-Augmented Generation in Olympic-Level Physics Problem Solving

Authors: Shunfeng Zheng, Yudi Zhang, Meng Fang, Zihan Zhang, Zhitan Wu, Mykola Pechenizkiy, Ling Chen
Published: 2025-10-01
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.00919v3

Retrieval-augmented generation (RAG) with foundation models has achieved strong performance across diverse tasks, but their capacity for expert-level reasoning-such as solving Olympiad-level physics problems-remains largely unexplored. Inspired by the way students prepare for competitions by reviewing past problems, we investigate the potential of RAG to enhance physics reasoning in foundation models. We introduce PhoPile, a high-quality multimodal dataset specifically designed for Olympiad-level physics, enabling systematic study of retrieval-based reasoning. PhoPile includes diagrams, graphs, and equations, capturing the inherently multimodal nature of physics problem solving. Using PhoPile, we benchmark RAG-augmented foundation models, covering both large language models (LLMs) and large multimodal models (LMMs) with multiple retrievers. Our results demonstrate that integrating retrieval with physics corpora can improve model performance, while also highlighting challenges that motivate further research in retrieval-augmented physics reasoning.

---

### Reinforcement Learning with Verifiable yet Noisy Rewards under Imperfect Verifiers

Authors: Xin-Qiang Cai, Wei Wang, Feng Liu, Tongliang Liu, Gang Niu, Masashi Sugiyama
Published: 2025-10-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.00915v5

Reinforcement Learning with Verifiable Rewards (RLVR) replaces costly human labeling with automated verifiers. To reduce verifier hacking, many RLVR systems binarize rewards to $\{0,1\}$, but imperfect verifiers inevitably introduce \emph{false negatives} (rejecting correct answers) and \emph{false positives} (accepting incorrect ones). We formalize verifier unreliability as a stochastic reward channel with asymmetric noise rates $ρ_0$ and $ρ_1$ -- the FP rate and the FN rate, respectively. From this abstraction we derive two lightweight corrections: (i) a \emph{backward} correction that yields an unbiased surrogate reward and thus an unbiased policy-gradient estimator in expectation, and (ii) a \emph{forward} correction that reweights score-function terms so the expected update aligns with the clean gradient direction and requires only the FN rate. We implement both as lightweight hooks in a group relative policy optimization pipeline, both corrections improve RLVR for math reasoning under synthetic and real verifier noise, with the forward variant being more stable under heavier noise. Finally, an appeals mechanism with a lightweight LLM verifier estimates the FN rate online and further improves performance.

---

### RiskPO: Risk-based Policy Optimization via Verifiable Reward for LLM Post-Training

Authors: Tao Ren, Jinyang Jiang, Hui Yang, Wan Tian, Minhao Zou, Guanghao Li, Zishi Zhang, Qinghao Wang, Shentao Qin, Yanjun Zhao, Rui Tao, Hui Shao et al.
Published: 2025-10-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.00911v1

Reinforcement learning with verifiable reward has recently emerged as a central paradigm for post-training large language models (LLMs); however, prevailing mean-based methods, such as Group Relative Policy Optimization (GRPO), suffer from entropy collapse and limited reasoning gains. We argue that these issues stem from overemphasizing high-probability output sequences while neglecting rare but informative reasoning paths. To address these challenges, we propose Risk-based Policy Optimization (RiskPO), which substitutes classical mean-based objectives with principled risk measures. Specifically, we introduce a Mixed Value-at-Risk objective that integrates weighted attention over multiple regions of the reward distribution, thereby amplifying gradient signals on challenging instances and preventing overconfident convergence. We further design a bundling scheme that aggregates multiple questions into bundles, thus enriching the feedback signal and yielding more stable and informative training dynamics. Theoretically, we prove that the risk-averse update alleviates entropy collapse and promotes exploration. Numerically, RiskPO achieves consistent and significant improvements in mathematical reasoning, multi-modal reasoning, and code generation benchmarks, surpassing GRPO and its variants on both Pass@1 and Pass@k metrics. Our results demonstrate that risk-based optimization provides a rigorous and effective paradigm for enhancing LLM reasoning capabilities.

---

### PAL-Net: A Point-Wise CNN with Patch-Attention for 3D Facial Landmark Localization

Authors: Ali Shadman Yazdi, Annalisa Cappella, Benedetta Baldini, Riccardo Solazzo, Gianluca Tartaglia, Chiarella Sforza, Giuseppe Baselli
Published: 2025-10-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.00910v2

Manual annotation of anatomical landmarks on 3D facial scans is a time-consuming and expertise-dependent task, yet it remains critical for clinical assessments, morphometric analysis, and craniofacial research. While several deep learning methods have been proposed for facial landmark localization, most focus on pseudo-landmarks or require complex input representations, limiting their clinical applicability. This study presents a fully automated deep learning pipeline (PAL-Net) for localizing 50 anatomical landmarks on stereo-photogrammetry facial models. The method combines coarse alignment, region-of-interest filtering, and an initial approximation of landmarks with a patch-based pointwise CNN enhanced by attention mechanisms. Trained and evaluated on 214 annotated scans from healthy adults, PAL-Net achieved a mean localization error of 3.686 mm and preserves relevant anatomical distances with a 2.822 mm average error, comparable to intra-observer variability. To assess generalization, the model was further evaluated on 700 subjects from the FaceScape dataset, achieving a point-wise error of 0.41\,mm and a distance-wise error of 0.38\,mm. Compared to existing methods, PAL-Net offers a favorable trade-off between accuracy and computational cost. While performance degrades in regions with poor mesh quality (e.g., ears, hairline), the method demonstrates consistent accuracy across most anatomical regions. PAL-Net generalizes effectively across datasets and facial regions, outperforming existing methods in both point-wise and structural evaluations. It provides a lightweight, scalable solution for high-throughput 3D anthropometric analysis, with potential to support clinical workflows and reduce reliance on manual annotation. Source code can be found at https://github.com/Ali5hadman/PAL-Net-A-Point-Wise-CNN-with-Patch-Attention

---

### "We are not Future-ready": Understanding AI Privacy Risks and Existing Mitigation Strategies from the Perspective of AI Developers in Europe

Authors: Alexandra Klymenko, Stephen Meisenbacher, Patrick Gage Kelley, Sai Teja Peddinti, Kurt Thomas, Florian Matthes
Published: 2025-10-01
Categories: cs.HC, cs.AI
arXiv: https://arxiv.org/abs/2510.00909v1

The proliferation of AI has sparked privacy concerns related to training data, model interfaces, downstream applications, and more. We interviewed 25 AI developers based in Europe to understand which privacy threats they believe pose the greatest risk to users, developers, and businesses and what protective strategies, if any, would help to mitigate them. We find that there is little consensus among AI developers on the relative ranking of privacy risks. These differences stem from salient reasoning patterns that often relate to human rather than purely technical factors. Furthermore, while AI developers are aware of proposed mitigation strategies for addressing these risks, they reported minimal real-world adoption. Our findings highlight both gaps and opportunities for empowering AI developers to better address privacy risks in AI.

---

### Bridging Language Gaps: Advances in Cross-Lingual Information Retrieval with Multilingual LLMs

Authors: Roksana Goworek, Olivia Macmillan-Scott, Eda B. Özyiğit
Published: 2025-10-01
Categories: cs.IR, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2510.00908v1

Cross-lingual information retrieval (CLIR) addresses the challenge of retrieving relevant documents written in languages different from that of the original query. Research in this area has typically framed the task as monolingual retrieval augmented by translation, treating retrieval methods and cross-lingual capabilities in isolation. Both monolingual and cross-lingual retrieval usually follow a pipeline of query expansion, ranking, re-ranking and, increasingly, question answering. Recent advances, however, have shifted from translation-based methods toward embedding-based approaches and leverage multilingual large language models (LLMs), for which aligning representations across languages remains a central challenge. The emergence of cross-lingual embeddings and multilingual LLMs has introduced a new paradigm, offering improved retrieval performance and enabling answer generation. This survey provides a comprehensive overview of developments from early translation-based methods to state-of-the-art embedding-driven and generative techniques. It presents a structured account of core CLIR components, evaluation practices, and available resources. Persistent challenges such as data imbalance and linguistic variation are identified, while promising directions are suggested for advancing equitable and effective cross-lingual information retrieval. By situating CLIR within the broader landscape of information retrieval and multilingual language processing, this work not only reviews current capabilities but also outlines future directions for building retrieval systems that are robust, inclusive, and adaptable.

---

### BoMGene: Integrating Boruta-mRMR feature selection for enhanced Gene expression classification

Authors: Bich-Chung Phan, Thanh Ma, Huu-Hoa Nguyen, Thanh-Nghi Do
Published: 2025-10-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.00907v1

Feature selection is a crucial step in analyzing gene expression data, enhancing classification performance, and reducing computational costs for high-dimensional datasets. This paper proposes BoMGene, a hybrid feature selection method that effectively integrates two popular techniques: Boruta and Minimum Redundancy Maximum Relevance (mRMR). The method aims to optimize the feature space and enhance classification accuracy. Experiments were conducted on 25 publicly available gene expression datasets, employing widely used classifiers such as Support Vector Machine (SVM), Random Forest, XGBoost (XGB), and Gradient Boosting Machine (GBM). The results show that using the Boruta-mRMR combination cuts down the number of features chosen compared to just using mRMR, which helps to speed up training time while keeping or even improving classification accuracy compared to using individual feature selection methods. The proposed approach demonstrates clear advantages in accuracy, stability, and practical applicability for multi-class gene expression data analysis

---

### TubeDAgger: Reducing the Number of Expert Interventions with Stochastic Reach-Tubes

Authors: Julian Lemmel, Manuel Kranzl, Adam Lamine, Philipp Neubauer, Radu Grosu, Sophie A. Neubauer
Published: 2025-10-01
Categories: eess.SY, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2510.00906v1

Interactive Imitation Learning deals with training a novice policy from expert demonstrations in an online fashion. The established DAgger algorithm trains a robust novice policy by alternating between interacting with the environment and retraining of the network. Many variants thereof exist, that differ in the method of discerning whether to allow the novice to act or return control to the expert. We propose the use of stochastic reachtubes - common in verification of dynamical systems - as a novel method for estimating the necessity of expert intervention. Our approach does not require fine-tuning of decision thresholds per environment and effectively reduces the number of expert interventions, especially when compared with related approaches that make use of a doubt classification model.

---

### Intuitions of Machine Learning Researchers about Transfer Learning for Medical Image Classification

Authors: Yucheng Lu, Hubert Dariusz Zając, Veronika Cheplygina, Amelia Jiménez-Sánchez
Published: 2025-10-01
Categories: cs.CV, cs.CY, cs.HC
arXiv: https://arxiv.org/abs/2510.00902v2

Transfer learning is crucial for medical imaging, yet the selection of source datasets often relies on researchers' intuition rather than systematic principles, which can impact the generalizability of algorithms and, thus, patient outcomes. This study investigates these decisions through a task-based survey with machine learning practitioners. Unlike prior work that benchmarks models and experimental setups, we take a human-computer interaction (HCI) perspective on how practitioners select source datasets. Our findings indicate that choices are task-dependent and influenced by community practices, dataset properties, and computational (data embedding), or perceived visual or semantic similarity. However, similarity ratings and expected performance are not always aligned, challenging a traditional "more similar is better" view. Moreover, ethical and fairness considerations remain largely absent from source dataset sections. Participants often used ambiguous terminology, which suggests a need for clearer definitions and tools to make them explicit and usable. By clarifying these heuristics and introducing a conceptual framework of transfer learning factors, this work provides practical insights for more systematic source selection in transfer learning.

---

### FusionAdapter for Few-Shot Relation Learning in Multimodal Knowledge Graphs

Authors: Ran Liu, Yuan Fang, Xiaoli Li
Published: 2025-10-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.00894v1

Multimodal Knowledge Graphs (MMKGs) incorporate various modalities, including text and images, to enhance entity and relation representations. Notably, different modalities for the same entity often present complementary and diverse information. However, existing MMKG methods primarily align modalities into a shared space, which tends to overlook the distinct contributions of specific modalities, limiting their performance particularly in low-resource settings. To address this challenge, we propose FusionAdapter for the learning of few-shot relationships (FSRL) in MMKG. FusionAdapter introduces (1) an adapter module that enables efficient adaptation of each modality to unseen relations and (2) a fusion strategy that integrates multimodal entity representations while preserving diverse modality-specific characteristics. By effectively adapting and fusing information from diverse modalities, FusionAdapter improves generalization to novel relations with minimal supervision. Extensive experiments on two benchmark MMKG datasets demonstrate that FusionAdapter achieves superior performance over state-of-the-art methods.

---

### Span-level Detection of AI-generated Scientific Text via Contrastive Learning and Structural Calibration

Authors: Zhen Yin, Shenghua Wang
Published: 2025-10-01
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.00890v1

The rapid adoption of large language models (LLMs) in scientific writing raises serious concerns regarding authorship integrity and the reliability of scholarly publications. Existing detection approaches mainly rely on document-level classification or surface-level statistical cues; however, they neglect fine-grained span localization, exhibit weak calibration, and often fail to generalize across disciplines and generators. To address these limitations, we present Sci-SpanDet, a structure-aware framework for detecting AI-generated scholarly texts. The proposed method combines section-conditioned stylistic modeling with multi-level contrastive learning to capture nuanced human-AI differences while mitigating topic dependence, thereby enhancing cross-domain robustness. In addition, it integrates BIO-CRF sequence labeling with pointer-based boundary decoding and confidence calibration to enable precise span-level detection and reliable probability estimates. Extensive experiments on a newly constructed cross-disciplinary dataset of 100,000 annotated samples generated by multiple LLM families (GPT, Qwen, DeepSeek, LLaMA) demonstrate that Sci-SpanDet achieves state-of-the-art performance, with F1(AI) of 80.17, AUROC of 92.63, and Span-F1 of 74.36. Furthermore, it shows strong resilience under adversarial rewriting and maintains balanced accuracy across IMRaD sections and diverse disciplines, substantially surpassing existing baselines. To ensure reproducibility and to foster further research on AI-generated text detection in scholarly documents, the curated dataset and source code will be publicly released upon publication.

---

### Rectifying Regression in Reinforcement Learning

Authors: Alex Ayoub, David Szepesvári, Alireza Bakhtiari, Csaba Szepesvári, Dale Schuurmans
Published: 2025-10-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.00885v2

This paper investigates the impact of the loss function in value-based methods for reinforcement learning through an analysis of underlying prediction objectives. We theoretically show that mean absolute error is a better prediction objective than the traditional mean squared error for controlling the learned policy's suboptimality gap. Furthermore, we present results that different loss functions are better aligned with these different regression objectives: binary and categorical cross-entropy losses with the mean absolute error and squared loss with the mean squared error. We then provide empirical evidence that algorithms minimizing these cross-entropy losses can outperform those based on the squared loss in linear reinforcement learning.

---

### COMMET: orders-of-magnitude speed-up in finite element method via batch-vectorized neural constitutive updates

Authors: Benjamin Alheit, Mathias Peirlinck, Siddhant Kumar
Published: 2025-10-01
Categories: cs.CE, cs.LG
arXiv: https://arxiv.org/abs/2510.00884v2

Constitutive evaluations often dominate the computational cost of finite element (FE) simulations whenever material models are complex. Neural constitutive models (NCMs) offer a highly expressive and flexible framework for modeling complex material behavior in solid mechanics. However, their practical adoption in large-scale FE simulations remains limited due to significant computational costs, especially in repeatedly evaluating stress and stiffness. NCMs thus represent an extreme case: their large computational graphs make stress and stiffness evaluations prohibitively expensive, restricting their use to small-scale problems. In this work, we introduce COMMET, an open-source FE framework whose architecture has been redesigned from the ground up to accelerate high-cost constitutive updates. Our framework features a novel assembly algorithm that supports batched and vectorized constitutive evaluations, compute-graph-optimized derivatives that replace automatic differentiation, and distributed-memory parallelism via MPI. These advances dramatically reduce runtime, with speed-ups exceeding three orders of magnitude relative to traditional non-vectorized automatic differentiation-based implementations. While we demonstrate these gains primarily for NCMs, the same principles apply broadly wherever for-loop based assembly or constitutive updates limit performance, establishing a new standard for large-scale, high-fidelity simulations in computational mechanics.

---

### GLAI: GreenLightningAI for Accelerated Training through Knowledge Decoupling

Authors: Jose I. Mestre, Alberto Fernández-Hernández, Cristian Pérez-Corral, Manuel F. Dolz, Jose Duato, Enrique S. Quintana-Ortí
Published: 2025-10-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.00883v2

In this work we introduce GreenLightningAI (GLAI), a new architectural block designed as an alternative to conventional MLPs. The central idea is to separate two types of knowledge that are usually entangled during training: (i) *structural knowledge*, encoded by the stable activation patterns induced by ReLU activations; and (ii) *quantitative knowledge*, carried by the numerical weights and biases. By fixing the structure once stabilized, GLAI reformulates the MLP as a combination of paths, where only the quantitative component is optimized. This reformulation retains the universal approximation capabilities of MLPs, yet achieves a more efficient training process, reducing training time by ~40% on average across the cases examined in this study. Crucially, GLAI is not just another classifier, but a generic block that can replace MLPs wherever they are used, from supervised heads with frozen backbones to projection layers in self-supervised learning or few-shot classifiers. Across diverse experimental setups, GLAI consistently matches or exceeds the accuracy of MLPs with an equivalent number of parameters, while converging faster. Overall, GLAI establishes a new design principle that opens a direction for future integration into large-scale architectures such as Transformers, where MLP blocks dominate the computational footprint.

---

### AI-CNet3D: An Anatomically-Informed Cross-Attention Network with Multi-Task Consistency Fine-tuning for 3D Glaucoma Classification

Authors: Roshan Kenia, Anfei Li, Rishabh Srivastava, Kaveri A. Thakoor
Published: 2025-10-01
Categories: cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2510.00882v1

Glaucoma is a progressive eye disease that leads to optic nerve damage, causing irreversible vision loss if left untreated. Optical coherence tomography (OCT) has become a crucial tool for glaucoma diagnosis, offering high-resolution 3D scans of the retina and optic nerve. However, the conventional practice of condensing information from 3D OCT volumes into 2D reports often results in the loss of key structural details. To address this, we propose a novel hybrid deep learning model that integrates cross-attention mechanisms into a 3D convolutional neural network (CNN), enabling the extraction of critical features from the superior and inferior hemiretinas, as well as from the optic nerve head (ONH) and macula, within OCT volumes. We introduce Channel Attention REpresentations (CAREs) to visualize cross-attention outputs and leverage them for consistency-based multi-task fine-tuning, aligning them with Gradient-Weighted Class Activation Maps (Grad-CAMs) from the CNN's final convolutional layer to enhance performance, interpretability, and anatomical coherence. We have named this model AI-CNet3D (AI-`See'-Net3D) to reflect its design as an Anatomically-Informed Cross-attention Network operating on 3D data. By dividing the volume along two axes and applying cross-attention, our model enhances glaucoma classification by capturing asymmetries between the hemiretinal regions while integrating information from the optic nerve head and macula. We validate our approach on two large datasets, showing that it outperforms state-of-the-art attention and convolutional models across all key metrics. Finally, our model is computationally efficient, reducing the parameter count by one-hundred--fold compared to other attention mechanisms while maintaining high diagnostic performance and comparable GFLOPS.

---

### Advancing Automated Ethical Profiling in SE: a Zero-Shot Evaluation of LLM Reasoning

Authors: Patrizio Migliarini, Mashal Afzal Memon, Marco Autili, Paola Inverardi
Published: 2025-10-01
Categories: cs.SE, cs.AI
arXiv: https://arxiv.org/abs/2510.00881v1

Large Language Models (LLMs) are increasingly integrated into software engineering (SE) tools for tasks that extend beyond code synthesis, including judgment under uncertainty and reasoning in ethically significant contexts. We present a fully automated framework for assessing ethical reasoning capabilities across 16 LLMs in a zero-shot setting, using 30 real-world ethically charged scenarios. Each model is prompted to identify the most applicable ethical theory to an action, assess its moral acceptability, and explain the reasoning behind their choice. Responses are compared against expert ethicists' choices using inter-model agreement metrics. Our results show that LLMs achieve an average Theory Consistency Rate (TCR) of 73.3% and Binary Agreement Rate (BAR) on moral acceptability of 86.7%, with interpretable divergences concentrated in ethically ambiguous cases. A qualitative analysis of free-text explanations reveals strong conceptual convergence across models despite surface-level lexical diversity. These findings support the potential viability of LLMs as ethical inference engines within SE pipelines, enabling scalable, auditable, and adaptive integration of user-aligned ethical reasoning. Our focus is the Ethical Interpreter component of a broader profiling pipeline: we evaluate whether current LLMs exhibit sufficient interpretive stability and theory-consistent reasoning to support automated profiling.

---

### HalluGuard: Evidence-Grounded Small Reasoning Models to Mitigate Hallucinations in Retrieval-Augmented Generation

Authors: Loris Bergeron, Ioana Buhnila, Jérôme François, Radu State
Published: 2025-10-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.00880v1

Large Language Models (LLMs) excel in many NLP tasks but remain prone to hallucinations, limiting trust in real-world applications. We present HalluGuard, a 4B-parameter Small Reasoning Model (SRM) for mitigating hallucinations in Retrieval-Augmented Generation (RAG). HalluGuard classifies document-claim pairs as grounded or hallucinated and produces evidence-grounded justifications for transparency. Our approach combines (i) a domain-agnostic synthetic dataset derived from FineWeb and refined through multi-stage curation and data reformation, (ii) synthetic grounded and hallucinated claims, and (iii) preference-based fine-tuning with Odds Ratio Preference Optimization to distill large-model reasoning into a smaller backbone. On the RAGTruth subset of the LLM-AggreFact benchmark, HalluGuard achieves 84.0% balanced accuracy (BAcc), rivaling specialized models, MiniCheck (7B; 84.0%) and Granite Guardian 3.3 (8B; 82.2%) while using roughly half their parameters. Over the full benchmark it reaches 75.7% BAcc, matching larger general-purpose LLMs such as GPT-4o (75.9%). We will release HalluGuard and datasets under Apache 2.0 upon acceptance.

---

### Unveiling Interesting Insights: Monte Carlo Tree Search for Knowledge Discovery

Authors: Pietro Totis, Alberto Pozanco, Daniel Borrajo
Published: 2025-10-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.00876v1

Organizations are increasingly focused on leveraging data from their processes to gain insights and drive decision-making. However, converting this data into actionable knowledge remains a difficult and time-consuming task. There is often a gap between the volume of data collected and the ability to process and understand it, which automated knowledge discovery aims to fill. Automated knowledge discovery involves complex open problems, including effectively navigating data, building models to extract implicit relationships, and considering subjective goals and knowledge. In this paper, we introduce a novel method for Automated Insights and Data Exploration (AIDE), that serves as a robust foundation for tackling these challenges through the use of Monte Carlo Tree Search (MCTS). We evaluate AIDE using both real-world and synthetic data, demonstrating its effectiveness in identifying data transformations and models that uncover interesting data patterns. Among its strengths, AIDE's MCTS-based framework offers significant extensibility, allowing for future integration of additional pattern extraction strategies and domain knowledge. This makes AIDE a valuable step towards developing a comprehensive solution for automated knowledge discovery.

---

### False Discovery Rate Control via Bayesian Mirror Statistic

Authors: Marco Molinari, Magne Thoresen
Published: 2025-10-01
Categories: stat.ME, stat.ML
arXiv: https://arxiv.org/abs/2510.00875v1

Simultaneously performing variable selection and inference in high-dimensional models is an open challenge in statistics and machine learning. The increasing availability of vast amounts of variables requires the adoption of specific statistical procedures to accurately select the most important predictors in a high-dimensional space, while being able to control some form of selection error. In this work we adapt the Mirror Statistic approach to False Discovery Rate (FDR) control into a Bayesian modelling framework. The Mirror Statistic, developed in the classic frequentist statistical framework, is a flexible method to control FDR, which only requires mild model assumptions, but requires two sets of independent regression coefficient estimates, usually obtained after splitting the original dataset. Here we propose to rely on a Bayesian formulation of the model and use the posterior distributions of the coefficients of interest to build the Mirror Statistic and effectively control the FDR without the need to split the data. Moreover, the method is very flexible since it can be used with continuous and discrete outcomes and more complex predictors, such as with mixed models. We keep the approach scalable to high-dimensions by relying on Automatic Differentiation Variational Inference and fully continuous prior choices.

---

### Reducción de ruido por medio de autoencoders: caso de estudio con la señal GW150914

Authors: Fernanda Zapata Bascuñán, Darío Fernando Mendieta
Published: 2025-10-01
Categories: cs.LG, astro-ph.IM
arXiv: https://arxiv.org/abs/2510.00873v1

This brief study focuses on the application of autoencoders to improve the quality of low-amplitude signals, such as gravitational events. A pre-existing autoencoder was trained using cosmic event data, optimizing its architecture and parameters. The results show a significant increase in the signal-to-noise ratio of the processed signals, demonstrating the potential of autoencoders in the analysis of small signals with multiple sources of interference.

---

### A Visual Diagnostics Framework for District Heating Data: Enhancing Data Quality for AI-Driven Heat Consumption Prediction

Authors: Kristoffer Christensen, Bo Nørregaard Jørgensen, Zheng Grace Ma
Published: 2025-10-01
Categories: cs.LG, cs.HC
arXiv: https://arxiv.org/abs/2510.00872v1

High-quality data is a prerequisite for training reliable Artificial Intelligence (AI) models in the energy domain. In district heating networks, sensor and metering data often suffer from noise, missing values, and temporal inconsistencies, which can significantly degrade model performance. This paper presents a systematic approach for evaluating and improving data quality using visual diagnostics, implemented through an interactive web-based dashboard. The dashboard employs Python-based visualization techniques, including time series plots, heatmaps, box plots, histograms, correlation matrices, and anomaly-sensitive KPIs such as skewness and anomaly detection based on the modified z-scores. These tools al-low human experts to inspect and interpret data anomalies, enabling a human-in-the-loop strategy for data quality assessment. The methodology is demonstrated on a real-world dataset from a Danish district heating provider, covering over four years of hourly data from nearly 7000 meters. The findings show how visual analytics can uncover systemic data issues and, in the future, guide data cleaning strategies that enhance the accuracy, stability, and generalizability of Long Short-Term Memory and Gated Recurrent Unit models for heat demand forecasting. The study contributes to a scalable, generalizable framework for visual data inspection and underlines the critical role of data quality in AI-driven energy management systems.

---

### Target Population Synthesis using CT-GAN

Authors: Tanay Rastogi, Daniel Jonsson
Published: 2025-10-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.00871v1

Agent-based models used in scenario planning for transportation and urban planning usually require detailed population information from the base as well as target scenarios. These populations are usually provided by synthesizing fake agents through deterministic population synthesis methods. However, these deterministic population synthesis methods face several challenges, such as handling high-dimensional data, scalability, and zero-cell issues, particularly when generating populations for target scenarios. This research looks into how a deep generative model called Conditional Tabular Generative Adversarial Network (CT-GAN) can be used to create target populations either directly from a collection of marginal constraints or through a hybrid method that combines CT-GAN with Fitness-based Synthesis Combinatorial Optimization (FBS-CO). The research evaluates the proposed population synthesis models against travel survey and zonal-level aggregated population data. Results indicate that the stand-alone CT-GAN model performs the best when compared with FBS-CO and the hybrid model. CT-GAN by itself can create realistic-looking groups that match single-variable distributions, but it struggles to maintain relationships between multiple variables. However, the hybrid model demonstrates improved performance compared to FBS-CO by leveraging CT-GAN ability to generate a descriptive base population, which is then refined using FBS-CO to align with target-year marginals. This study demonstrates that CT-GAN represents an effective methodology for target populations and highlights how deep generative models can be successfully integrated with conventional synthesis techniques to enhance their performance.

---

### Removing Noise, not Finding Gold: Quality Filtering for Large-Scale Pretraining

Authors: Thiziri Nait Saada, Louis Bethune, Michal Klein, David Grangier, Marco Cuturi, Pierre Ablin
Published: 2025-10-01
Categories: cs.LG, cs.CL
arXiv: https://arxiv.org/abs/2510.00866v4

Large-scale models are pretrained on massive web-crawled datasets containing documents of mixed quality, making data filtering essential. A popular method is Classifier-based Quality Filtering (CQF), which trains a binary classifier to distinguish between pretraining data and a small, high-quality set. It assigns each pretraining document a quality score defined as the classifier's score and retains only the top-scoring ones. We provide an in-depth analysis of CQF. We show that while CQF improves downstream task performance, it does not necessarily enhance language modeling on the high-quality set. Importantly, we find that training on CQF-selected data can outperform training directly on the high-quality set, even when the latter is sufficiently large. This finding alone is particularly striking, given the substantial effort and cost recently devoted to augmenting high-quality data. We explain this paradox by the fact that CQF implicitly filters the high-quality dataset as well as the low-quality one. Finally, we introduce an optimization-driven notion of data quality and demonstrate that it can be reliably estimated using small-scale proxy experiments. Altogether, our results both elucidate the mechanisms behind CQF and deepen our understanding of data selection methods widely used in practice.

---

### Gather-Scatter Mamba: Accelerating Propagation with Efficient State Space Model

Authors: Hyun-kyu Ko, Youbin Kim, Jihyeon Park, Dongheok Park, Gyeongjin Kang, Wonjun Cho, Hyung Yi, Eunbyung Park
Published: 2025-10-01
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2510.00862v1

State Space Models (SSMs)-most notably RNNs-have historically played a central role in sequential modeling. Although attention mechanisms such as Transformers have since dominated due to their ability to model global context, their quadratic complexity and limited scalability make them less suited for long sequences. Video super-resolution (VSR) methods have traditionally relied on recurrent architectures to propagate features across frames. However, such approaches suffer from well-known issues including vanishing gradients, lack of parallelism, and slow inference speed. Recent advances in selective SSMs like Mamba offer a compelling alternative: by enabling input-dependent state transitions with linear-time complexity, Mamba mitigates these issues while maintaining strong long-range modeling capabilities. Despite this potential, Mamba alone struggles to capture fine-grained spatial dependencies due to its causal nature and lack of explicit context aggregation. To address this, we propose a hybrid architecture that combines shifted window self-attention for spatial context aggregation with Mamba-based selective scanning for efficient temporal propagation. Furthermore, we introduce Gather-Scatter Mamba (GSM), an alignment-aware mechanism that warps features toward a center anchor frame within the temporal window before Mamba propagation and scatters them back afterward, effectively reducing occlusion artifacts and ensuring effective redistribution of aggregated information across all frames. The official implementation is provided at: https://github.com/Ko-Lani/GSMamba.

---

### Erase to Improve: Erasable Reinforcement Learning for Search-Augmented LLMs

Authors: Ziliang Wang, Kang An, Xuhui Zheng, Faqiang Qian, Weikun Zhang, Cijun Ouyang, Jialu Cai, Yuhang Wang, Yichao Wu
Published: 2025-10-01
Categories: cs.CL, cs.AI, cs.IR
arXiv: https://arxiv.org/abs/2510.00861v2

While search-augmented large language models (LLMs) exhibit impressive capabilities, their reliability in complex multi-hop reasoning remains limited. This limitation arises from three fundamental challenges: decomposition errors, where tasks are incorrectly broken down; retrieval missing, where key evidence fails to be retrieved; and reasoning errors, where flawed logic propagates through the reasoning chain. A single failure in any of these stages can derail the final answer. We propose Erasable Reinforcement Learning (ERL), a novel framework that transforms fragile reasoning into a robust process. ERL explicitly identifies faulty steps, erases them, and regenerates reasoning in place, preventing defective logic from propagating through the reasoning chain. This targeted correction mechanism turns brittle reasoning into a more resilient process. Models trained with ERL, termed ESearch, achieve substantial improvements on HotpotQA, MuSiQue, 2Wiki, and Bamboogle, with the 3B model achieving +8.48% EM and +11.56% F1, and the 7B model achieving +5.38% EM and +7.22% F1 over previous state-of-the-art(SOTA) results. These findings suggest that erasable reinforcement learning provides a powerful paradigm shift for robust multi-step reasoning in LLMs.

---

### Population Synthesis using Incomplete Information

Authors: Tanay Rastogi, Daniel Jonsson, Anders Karlström
Published: 2025-10-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.00859v1

This paper presents a population synthesis model that utilizes the Wasserstein Generative-Adversarial Network (WGAN) for training on incomplete microsamples. By using a mask matrix to represent missing values, the study proposes a WGAN training algorithm that lets the model learn from a training dataset that has some missing information. The proposed method aims to address the challenge of missing information in microsamples on one or more attributes due to privacy concerns or data collection constraints. The paper contrasts WGAN models trained on incomplete microsamples with those trained on complete microsamples, creating a synthetic population. We conducted a series of evaluations of the proposed method using a Swedish national travel survey. We validate the efficacy of the proposed method by generating synthetic populations from all the models and comparing them to the actual population dataset. The results from the experiments showed that the proposed methodology successfully generates synthetic data that closely resembles a model trained with complete data as well as the actual population. The paper contributes to the field by providing a robust solution for population synthesis with incomplete data, opening avenues for future research, and highlighting the potential of deep generative models in advancing population synthesis capabilities.

---

### ManagerBench: Evaluating the Safety-Pragmatism Trade-off in Autonomous LLMs

Authors: Adi Simhi, Jonathan Herzig, Martin Tutek, Itay Itzhak, Idan Szpektor, Yonatan Belinkov
Published: 2025-10-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.00857v2

As large language models (LLMs) evolve from conversational assistants into autonomous agents, evaluating the safety of their actions becomes critical. Prior safety benchmarks have primarily focused on preventing generation of harmful content, such as toxic text. However, they overlook the challenge of agents taking harmful actions when the most effective path to an operational goal conflicts with human safety. To address this gap, we introduce ManagerBench, a benchmark that evaluates LLM decision-making in realistic, human-validated managerial scenarios. Each scenario forces a choice between a pragmatic but harmful action that achieves an operational goal, and a safe action that leads to worse operational performance. A parallel control set, where potential harm is directed only at inanimate objects, measures a model's pragmatism and identifies its tendency to be overly safe. Our findings indicate that the frontier LLMs perform poorly when navigating this safety-pragmatism trade-off. Many consistently choose harmful options to advance their operational goals, while others avoid harm only to become overly safe and ineffective. Critically, we find this misalignment does not stem from an inability to perceive harm, as models' harm assessments align with human judgments, but from flawed prioritization. ManagerBench is a challenging benchmark for a core component of agentic behavior: making safe choices when operational goals and alignment values incentivize conflicting actions. Benchmark & code available at https://technion-cs-nlp.github.io/ManagerBench-website/.

---

### Can World Models Benefit VLMs for World Dynamics?

Authors: Kevin Zhang, Kuangzhi Ge, Xiaowei Chi, Renrui Zhang, Shaojun Shi, Zhen Dong, Sirui Han, Shanghang Zhang
Published: 2025-10-01
Categories: cs.CV, cs.AI, cs.CL, cs.LG
arXiv: https://arxiv.org/abs/2510.00855v1

Trained on internet-scale video data, generative world models are increasingly recognized as powerful world simulators that can generate consistent and plausible dynamics over structure, motion, and physics. This raises a natural question: with the advent of strong video foundational models, might they supplant conventional vision encoder paradigms for general-purpose multimodal understanding? While recent studies have begun to explore the potential of world models on common vision tasks, these explorations typically lack a systematic investigation of generic, multimodal tasks. In this work, we strive to investigate the capabilities when world model priors are transferred into Vision-Language Models: we re-purpose a video diffusion model as a generative encoder to perform a single denoising step and treat the resulting latents as a set of visual embedding. We empirically investigate this class of models, which we refer to as World-Language Models (WorldLMs), and we find that generative encoders can capture latents useful for downstream understanding that show distinctions from conventional encoders. Naming our best-performing variant Dynamic Vision Aligner (DyVA), we further discover that this method significantly enhances spatial reasoning abilities and enables single-image models to perform multi-frame reasoning. Through the curation of a suite of visual reasoning tasks, we find DyVA to surpass both open-source and proprietary baselines, achieving state-of-the-art or comparable performance. We attribute these gains to WorldLM's inherited motion-consistency internalization from video pre-training. Finally, we systematically explore extensive model designs to highlight promising directions for future work. We hope our study can pave the way for a new family of VLMs that leverage priors from world models and are on a promising path towards generalist vision learners.

---

### Mechanistic Interpretability as Statistical Estimation: A Variance Analysis

Authors: Maxime Méloux, François Portet, Maxime Peyrard
Published: 2025-10-01
Categories: cs.LG, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2510.00845v4

Mechanistic Interpretability (MI) aims to reverse-engineer model behaviors by identifying functional sub-networks. Yet, the scientific validity of these findings depends on their stability. In this work, we argue that circuit discovery is not a standalone task but a statistical estimation problem built upon causal mediation analysis (CMA). We uncover a fundamental instability at this base layer: exact, single-input CMA scores exhibit high intrinsic variance, implying that the causal effect of a component is a volatile random variable rather than a fixed property. We then demonstrate that circuit discovery pipelines inherit this variance and further amplify it. Fast approximation methods, such as Edge Attribution Patching and its successors, introduce additional estimation noise, while aggregating these noisy scores over datasets leads to fragile structural estimates. Consequently, small perturbations in input data or hyperparameters yield vastly different circuits. We systematically decompose these sources of variance and advocate for more rigorous MI practices, prioritizing statistical robustness and routine reporting of stability metrics.

---

### Learning Compact Representations of LLM Abilities via Item Response Theory

Authors: Jianhao Chen, Chenxu Wang, Gengrui Zhang, Peng Ye, Lei Bai, Wei Hu, Yuzhong Qu, Shuyue Hu
Published: 2025-10-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.00844v1

Recent years have witnessed a surge in the number of large language models (LLMs), yet efficiently managing and utilizing these vast resources remains a significant challenge. In this work, we explore how to learn compact representations of LLM abilities that can facilitate downstream tasks, such as model routing and performance prediction on new benchmarks. We frame this problem as estimating the probability that a given model will correctly answer a specific query. Inspired by the item response theory (IRT) in psychometrics, we model this probability as a function of three key factors: (i) the model's multi-skill ability vector, (2) the query's discrimination vector that separates models of differing skills, and (3) the query's difficulty scalar. To learn these parameters jointly, we introduce a Mixture-of-Experts (MoE) network that couples model- and query-level embeddings. Extensive experiments demonstrate that our approach leads to state-of-the-art performance in both model routing and benchmark accuracy prediction. Moreover, analysis validates that the learned parameters encode meaningful, interpretable information about model capabilities and query characteristics.

---

### LLM Routing with Dueling Feedback

Authors: Chao-Kai Chiang, Takashi Ishida, Masashi Sugiyama
Published: 2025-10-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.00841v1

We study LLM routing, the problem of selecting the best model for each query while balancing user satisfaction, model expertise, and inference cost. We formulate routing as contextual dueling bandits, learning from pairwise preference feedback rather than absolute scores, thereby yielding label-efficient and dynamic adaptation. Building on this formulation, we introduce Category-Calibrated Fine-Tuning (CCFT), a representation-learning method that derives model embeddings from offline data using contrastive fine-tuning with categorical weighting. These embeddings enable the practical instantiation of Feel-Good Thompson Sampling for Contextual Dueling Bandits (FGTS.CDB), a theoretically grounded posterior-sampling algorithm. We propose four variants of the categorical weighting that explicitly integrate model quality and cost, and we empirically evaluate the proposed methods on the RouterBench and MixInstruct datasets. Across both benchmarks, our methods achieve lower cumulative regret and faster convergence, with better robustness and performance-cost balance than strong baselines built with a general-purpose OpenAI embedding model.

---

### Feature Identification for Hierarchical Contrastive Learning

Authors: Julius Ott, Nastassia Vysotskaya, Huawei Sun, Lorenzo Servadei, Robert Wille
Published: 2025-10-01
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2510.00837v1

Hierarchical classification is a crucial task in many applications, where objects are organized into multiple levels of categories. However, conventional classification approaches often neglect inherent inter-class relationships at different hierarchy levels, thus missing important supervisory signals. Thus, we propose two novel hierarchical contrastive learning (HMLC) methods. The first, leverages a Gaussian Mixture Model (G-HMLC) and the second uses an attention mechanism to capture hierarchy-specific features (A-HMLC), imitating human processing. Our approach explicitly models inter-class relationships and imbalanced class distribution at higher hierarchy levels, enabling fine-grained clustering across all hierarchy levels. On the competitive CIFAR100 and ModelNet40 datasets, our method achieves state-of-the-art performance in linear evaluation, outperforming existing hierarchical contrastive learning methods by 2 percentage points in terms of accuracy. The effectiveness of our approach is backed by both quantitative and qualitative results, highlighting its potential for applications in computer vision and beyond.

---

### Improving Cryptocurrency Pump-and-Dump Detection through Ensemble-Based Models and Synthetic Oversampling Techniques

Authors: Jieun Yu, Minjung Park, Sangmi Chai
Published: 2025-10-01
Categories: cs.AI, cs.CE, q-fin.RM
arXiv: https://arxiv.org/abs/2510.00836v1

This study aims to detect pump and dump (P&D) manipulation in cryptocurrency markets, where the scarcity of such events causes severe class imbalance and hinders accurate detection. To address this issue, the Synthetic Minority Oversampling Technique (SMOTE) was applied, and advanced ensemble learning models were evaluated to distinguish manipulative trading behavior from normal market activity. The experimental results show that applying SMOTE greatly enhanced the ability of all models to detect P&D events by increasing recall and improving the overall balance between precision and recall. In particular, XGBoost and LightGBM achieved high recall rates (94.87% and 93.59%, respectively) with strong F1-scores and demonstrated fast computational performance, making them suitable for near real time surveillance. These findings indicate that integrating data balancing techniques with ensemble methods significantly improves the early detection of manipulative activities, contributing to a fairer, more transparent, and more stable cryptocurrency market.

---

### Towards Verifiable Federated Unlearning: Framework, Challenges, and The Road Ahead

Authors: Thanh Linh Nguyen, Marcela Tuler de Oliveira, An Braeken, Aaron Yi Ding, Quoc-Viet Pham
Published: 2025-10-01
Categories: cs.DC, cs.AI
arXiv: https://arxiv.org/abs/2510.00833v2

Federated unlearning (FUL) enables removing the data influence from the model trained across distributed clients, upholding the right to be forgotten as mandated by privacy regulations. FUL facilitates a value exchange where clients gain privacy-preserving control over their data contributions, while service providers leverage decentralized computing and data freshness. However, this entire proposition is undermined because clients have no reliable way to verify that their data influence has been provably removed, as current metrics and simple notifications offer insufficient assurance. We envision unlearning verification becoming a pivotal and trust-by-design part of the FUL life-cycle development, essential for highly regulated and data-sensitive services and applications like healthcare. This article introduces veriFUL, a reference framework for verifiable FUL that formalizes verification entities, goals, approaches, and metrics. Specifically, we consolidate existing efforts and contribute new insights, concepts, and metrics to this domain. Finally, we highlight research challenges and identify potential applications and developments for verifiable FUL and veriFUL. This article aims to provide a comprehensive resource for researchers and practitioners to navigate and advance the field of verifiable FUL.

---

### Controlled Comparison of Machine Learning Models for Fault Classification and Localization in Power System Protection

Authors: Julian Oelhaf, Georg Kordowich, Changhun Kim, Paula Andrea Pérez-Toro, Christian Bergler, Andreas Maier, Johann Jäger, Siming Bayer
Published: 2025-10-01
Categories: cs.AI, cs.LG, eess.SP
arXiv: https://arxiv.org/abs/2510.00831v2

The increasing complexity of modern power systems, driven by the integration of inverter-based and distributed energy resources, challenges the reliability of conventional protection schemes and motivates the use of machine learning for protection tasks. However, published results are often difficult to compare because datasets, sensing assumptions, and decision horizons vary across studies. This paper presents a controlled comparison of machine learning models for fault classification (FC) and fault localization (FL) under identical sensing, timing, and validation conditions on a common electromagnetic transient dataset, using decision windows of 10-50 ms to reflect protection-relevant time scales. For FC, the best-performing nonlinear models achieve F1 scores above 0.98 already at 10 ms, while lower-capacity models degrade at shorter horizons but improve with longer windows, indicating that relevant fault-type information is already present in the earliest transient. For FL, the top-performing models reach a stable localization error of about 10 % of normalized line length across all evaluated horizons, while weaker models form a clearly separated second performance tier. Line-resolved analysis shows that localization accuracy varies across grid segments, indicating topology-dependent difficulty rather than insufficient temporal context alone. These findings provide a controlled reference for comparing machine learning models across two protection tasks with fundamentally different information requirements.

---

### Exposing the Cracks: Vulnerabilities of Retrieval-Augmented LLM-based Machine Translation

Authors: Yanming Sun, Runzhe Zhan, Chi Seng Cheang, Han Wu, Xuebo Liu, Yuyao Niu, Fengying Ye, Kaixin Lan, Lidia S. Chao, Derek F. Wong
Published: 2025-10-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.00829v2

\textbf{RE}trieval-\textbf{A}ugmented \textbf{L}LM-based \textbf{M}achine \textbf{T}ranslation (REAL-MT) shows promise for knowledge-intensive tasks like idiomatic translation, but its reliability under noisy retrieval contexts remains poorly understood despite this being a common challenge in real-world deployment. To address this gap, we propose a noise synthesis framework and new metrics to evaluate the robustness of REAL-MT systematically. Using this framework, we instantiate REAL-MT with Qwen-series models, including standard LLMs and large reasoning models (LRMs) with enhanced reasoning, and evaluate their performance on idiomatic translation across high-, medium-, and low-resource language pairs under synthesized noise. Our results show that low-resource language pairs, which rely more heavily on retrieved context, degrade more severely under noise than high-resource ones and often produce nonsensical translations. Although LRMs possess enhanced reasoning capabilities, they show no improvement in error correction and are even more susceptible to noise, tending to rationalize incorrect contexts. We find that this stems from an attention shift away from the source idiom to noisy content, while confidence increases despite declining accuracy, indicating poor calibration. To mitigate these issues, we investigate training-free and fine-tuning strategies, which improve robustness at the cost of performance in clean contexts, revealing a fundamental trade-off. Our findings highlight the limitations of current approaches, underscoring the need for self-verifying integration mechanisms.

---

### Non-Euclidean Broximal Point Method: A Blueprint for Geometry-Aware Optimization

Authors: Kaja Gruntkowska, Peter Richtárik
Published: 2025-10-01
Categories: math.OC, cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2510.00823v1

The recently proposed Broximal Point Method (BPM) (Gruntkowska et al., 2025) offers an idealized optimization framework based on iteratively minimizing the objective function over norm balls centered at the current iterate. It enjoys striking global convergence guarantees, converging linearly and in a finite number of steps for proper, closed and convex functions. However, its theoretical analysis has so far been confined to the Euclidean geometry. At the same time, emerging trends in deep learning optimization, exemplified by algorithms such as Muon (Jordan et al., 2024) and Scion (Pethick et al., 2025), demonstrate the practical advantages of minimizing over balls defined via non-Euclidean norms which better align with the underlying geometry of the associated loss landscapes. In this note, we ask whether the convergence theory of BPM can be extended to this more general, non-Euclidean setting. We give a positive answer, showing that most of the elegant guarantees of the original method carry over to arbitrary norm geometries. Along the way, we clarify which properties are preserved and which necessarily break down when leaving the Euclidean realm. Our analysis positions Non-Euclidean BPM as a conceptual blueprint for understanding a broad class of geometry-aware optimization algorithms, shedding light on the principles behind their practical effectiveness.

---

### Logical Consistency Between Disagreeing Experts and Its Role in AI Safety

Authors: Andrés Corrada-Emmanuel
Published: 2025-10-01
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.00821v1

If two experts disagree on a test, we may conclude both cannot be 100 per cent correct. But if they completely agree, no possible evaluation can be excluded. This asymmetry in the utility of agreements versus disagreements is explored here by formalizing a logic of unsupervised evaluation for classifiers. Its core problem is computing the set of group evaluations that are logically consistent with how we observe them agreeing and disagreeing in their decisions. Statistical summaries of their aligned decisions are inputs into a Linear Programming problem in the integer space of possible correct or incorrect responses given true labels. Obvious logical constraints, such as, the number of correct responses cannot exceed the number of observed responses, are inequalities. But in addition, there are axioms, universally applicable linear equalities that apply to all finite tests. The practical and immediate utility of this approach to unsupervised evaluation using only logical consistency is demonstrated by building no-knowledge alarms that can detect when one or more LLMs-as-Judges are violating a minimum grading threshold specified by the user.

---

### NSARM: Next-Scale Autoregressive Modeling for Robust Real-World Image Super-Resolution

Authors: Xiangtao Kong, Rongyuan Wu, Shuaizheng Liu, Lingchen Sun, Lei Zhang
Published: 2025-10-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.00820v1

Most recent real-world image super-resolution (Real-ISR) methods employ pre-trained text-to-image (T2I) diffusion models to synthesize the high-quality image either from random Gaussian noise, which yields realistic results but is slow due to iterative denoising, or directly from the input low-quality image, which is efficient but at the price of lower output quality. These approaches train ControlNet or LoRA modules while keeping the pre-trained model fixed, which often introduces over-enhanced artifacts and hallucinations, suffering from the robustness to inputs of varying degradations. Recent visual autoregressive (AR) models, such as pre-trained Infinity, can provide strong T2I generation capabilities while offering superior efficiency by using the bitwise next-scale prediction strategy. Building upon next-scale prediction, we introduce a robust Real-ISR framework, namely Next-Scale Autoregressive Modeling (NSARM). Specifically, we train NSARM in two stages: a transformation network is first trained to map the input low-quality image to preliminary scales, followed by an end-to-end full-model fine-tuning. Such a comprehensive fine-tuning enhances the robustness of NSARM in Real-ISR tasks without compromising its generative capability. Extensive quantitative and qualitative evaluations demonstrate that as a pure AR model, NSARM achieves superior visual results over existing Real-ISR methods while maintaining a fast inference speed. Most importantly, it demonstrates much higher robustness to the quality of input images, showing stronger generalization performance. Project page: https://github.com/Xiangtaokong/NSARM

---

### Stabilizing Policy Gradients for Sample-Efficient Reinforcement Learning in LLM Reasoning

Authors: Luckeciano C. Melo, Alessandro Abate, Yarin Gal
Published: 2025-10-01
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.00819v2

Reinforcement Learning, particularly through policy gradient methods, has played a central role in enabling reasoning capabilities of Large Language Models. However, the optimization stability of policy gradients in this setting remains understudied. As a result, existing implementations often resort to conservative hyperparameter choices to ensure stability, which requires more training samples and increases computational costs. Hence, developing models for reliably tracking the underlying optimization dynamics and leveraging them into training enables more sample-efficient regimes and further unleashes scalable post-training. We address this gap by formalizing the stochastic optimization problem of policy gradients with explicit consideration of second-order geometry. We propose a tractable computational framework that tracks and leverages curvature information during policy updates. We further employ this framework to design interventions in the optimization process through data selection. The resultant algorithm, Curvature-Aware Policy Optimization (CAPO), identifies samples that contribute to unstable updates and masks them out. Theoretically, we establish monotonic improvement guarantees under realistic assumptions. On standard math reasoning benchmarks, we empirically show that CAPO ensures stable updates under aggressive learning regimes where baselines catastrophically fail. With minimal intervention (rejecting fewer than 8% of tokens), CAPO achieves up to 30x improvement in sample efficiency over standard GRPO for LLM reasoning.

---

### PhraseStereo: The First Open-Vocabulary Stereo Image Segmentation Dataset

Authors: Thomas Campagnolo, Ezio Malis, Philippe Martinet, Gaetan Bahl
Published: 2025-10-01
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.00818v1

Understanding how natural language phrases correspond to specific regions in images is a key challenge in multimodal semantic segmentation. Recent advances in phrase grounding are largely limited to single-view images, neglecting the rich geometric cues available in stereo vision. For this, we introduce PhraseStereo, the first novel dataset that brings phrase-region segmentation to stereo image pairs. PhraseStereo builds upon the PhraseCut dataset by leveraging GenStereo to generate accurate right-view images from existing single-view data, enabling the extension of phrase grounding into the stereo domain. This new setting introduces unique challenges and opportunities for multimodal learning, particularly in leveraging depth cues for more precise and context-aware grounding. By providing stereo image pairs with aligned segmentation masks and phrase annotations, PhraseStereo lays the foundation for future research at the intersection of language, vision, and 3D perception, encouraging the development of models that can reason jointly over semantics and geometry. The PhraseStereo dataset will be released online upon acceptance of this work.

---

### Semantic Bridges Between First Order c-Representations and Cost-Based Semantics: An Initial Perspective

Authors: Nicholas Leisegang, Giovanni Casini, Thomas Meyer
Published: 2025-10-01
Categories: cs.AI, cs.LO
arXiv: https://arxiv.org/abs/2510.00817v2

Weighted-knowledge bases and cost-based semantics represent a recent formalism introduced by Bienvenu et al. for Ontology Mediated Data Querying in the case where a given knowledge base is inconsistent. This is done by adding a weight to each statement in the knowledge base (KB), and then giving each DL interpretation a cost based on how often it breaks rules in the KB. In this paper we compare this approach with c-representations, a form of non-monotonic reasoning originally introduced by Kern-Isberner. c-Representations describe a means to interpret defeasible concept inclusions in the first-order case. This is done by assigning a numerical ranking to each interpretations via penalties for each violated conditional. We compare these two approaches on a semantic level. In particular, we show that under certain conditions a weighted knowledge base and a set of defeasible conditionals can generate the same ordering on interpretations, and therefore an equivalence of semantic structures up to relative cost. Moreover, we compare entailment described in both cases, where certain notions are equivalently expressible in both formalisms. Our results have the potential to benefit further work on both cost-based semantics and c-representations

---

### Learn to Guide Your Diffusion Model

Authors: Alexandre Galashov, Ashwini Pokle, Arnaud Doucet, Arthur Gretton, Mauricio Delbracio, Valentin De Bortoli
Published: 2025-10-01
Categories: cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2510.00815v1

Classifier-free guidance (CFG) is a widely used technique for improving the perceptual quality of samples from conditional diffusion models. It operates by linearly combining conditional and unconditional score estimates using a guidance weight $ω$. While a large, static weight can markedly improve visual results, this often comes at the cost of poorer distributional alignment. In order to better approximate the target conditional distribution, we instead learn guidance weights $ω_{c,(s,t)}$, which are continuous functions of the conditioning $c$, the time $t$ from which we denoise, and the time $s$ towards which we denoise. We achieve this by minimizing the distributional mismatch between noised samples from the true conditional distribution and samples from the guided diffusion process. We extend our framework to reward guided sampling, enabling the model to target distributions tilted by a reward function $R(x_0,c)$, defined on clean data and a conditioning $c$. We demonstrate the effectiveness of our methodology on low-dimensional toy examples and high-dimensional image settings, where we observe improvements in Fréchet inception distance (FID) for image generation. In text-to-image applications, we observe that employing a reward function given by the CLIP score leads to guidance weights that improve image-prompt alignment.

---

### Family Matters: Language Transfer and Merging for Adapting Small LLMs to Faroese

Authors: Jenny Kunz, Iben Nyholm Debess, Annika Simonsen
Published: 2025-10-01
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.00810v2

We investigate strategies for adapting small, efficient language models to Faroese, a low-resource North Germanic language. Starting from English-pretrained models, we apply continued pre-training on related Scandinavian languages -- individually or combined via model merging -- before fine-tuning on Faroese. We compare full fine-tuning with parameter-efficient adaptation via LoRA, assessing their effects on general language modeling performance, linguistic accuracy, and text comprehension. To address the lack of existing Faroese evaluation resources, we construct two new minimal-pair probing benchmarks, one for linguistic acceptability and one for text comprehension, and complement them with human evaluations conducted by native Faroese linguists. Our results show that transfer from related languages is essential, but the optimal source language is task-dependent: Icelandic improves linguistic accuracy, while Danish boosts reading comprehension. The choice of adaptation method likewise depends on the target task: LoRA yields stronger linguistic acceptability and marginally higher human evaluation scores, whereas full fine-tuning produces better comprehension performance and more robust downstream fine-tuning. Merging multiple related languages under full fine-tuning (but not LoRA) improves general language modeling, though its benefits in the linguistic acceptability and comprehension probes are less consistent.

---

### Foundation vs. Specialized Models: Evaluating Catastrophic Forgetting in Continual Time Series Forecasting

Authors: Nouha Karaouli, Denis Coquenet, Elisa Fromont, Martial Mermillod, Marina Reyboz
Published: 2025-10-01
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.00809v3

While Time Series Foundation Models (TSFMs) excel in zero-shot tasks, their behavior under continual fine tuning is poorly understood. We present the first systematic study of catastrophic forgetting in TSFMs (TimesFM-2.0, Chronos-2) versus a specialized SamFormer model across synthetic and real-world energy forecasting benchmarks. Our results show that while fine-tuning improves new task accuracy, it consistently triggers forgetting, though larger models exhibit greater inherent robustness. Notably, employing forgetting mitigation techniques such as DER, levels the playing field: it provides disproportionate gains to smaller models, allowing them to match TSFM performance by the end of the continual learning sequence. These findings suggest that in realistic, non-stationary scenarios, the high computational cost of large foundation models may not be justified over smaller models equipped with effective mitigation strategies.

---

### What You See is What You Ask: Evaluating Audio Descriptions

Authors: Divy Kala, Eshika Khandelwal, Makarand Tapaswi
Published: 2025-10-01
Categories: cs.CV, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2510.00808v1

Audio descriptions (ADs) narrate important visual details in movies, enabling Blind and Low Vision (BLV) users to understand narratives and appreciate visual details. Existing works in automatic AD generation mostly focus on few-second trimmed clips, and evaluate them by comparing against a single ground-truth reference AD. However, writing ADs is inherently subjective. Through alignment and analysis of two independent AD tracks for the same movies, we quantify the subjectivity in when and whether to describe, and what and how to highlight. Thus, we show that working with trimmed clips is inadequate. We propose ADQA, a QA benchmark that evaluates ADs at the level of few-minute long, coherent video segments, testing whether they would help BLV users understand the story and appreciate visual details. ADQA features visual appreciation (VA) questions about visual facts and narrative understanding (NU) questions based on the plot. Through ADQA, we show that current AD generation methods lag far behind human-authored ADs. We conclude with several recommendations for future work and introduce a public leaderboard for benchmarking.

---