# Research Papers — 2025-10 · bundle page 123 of 217

> Full-text markdown of papers #6101–6150 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/123
[← Previous bundle](https://aipod.dev/papers/archive/2025-10/page/122.md)
[Next bundle →](https://aipod.dev/papers/archive/2025-10/page/124.md)

### Flows, straight but not so fast: Exploring the design space of Rectified Flows in Protein Design

Authors: Junhua Chen, Simon Mathis, Charles Harris, Kieran Didi, Pietro Lio
Published: 2025-10-13
Categories: q-bio.BM, cs.AI
arXiv: https://arxiv.org/abs/2510.24732v1

Generative modeling techniques such as Diffusion and Flow Matching have achieved significant successes in generating designable and diverse protein backbones. However, many current models are computationally expensive, requiring hundreds or even thousands of function evaluations (NFEs) to yield samples of acceptable quality, which can become a bottleneck in practical design campaigns that often generate $10^4\ -\ 10^6$ designs per target. In image generation, Rectified Flows (ReFlow) can significantly reduce the required NFEs for a given target quality, but their application in protein backbone generation has been less studied. We apply ReFlow to improve the low NFE performance of pretrained SE(3) flow matching models for protein backbone generation and systematically study ReFlow design choices in the context of protein generation in data curation, training and inference time settings. In particular, we (1) show that ReFlow in the protein domain is particularly sensitive to the choice of coupling generation and annealing, (2) demonstrate how useful design choices for ReFlow in the image domain do not directly translate to better performance on proteins, and (3) make improvements to ReFlow methodology for proteins.

---

### Proportion and Perspective Control for Flow-Based Image Generation

Authors: Julien Boudier, Hugo Caselles-Dupré
Published: 2025-10-13
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2510.21763v1

While modern text-to-image diffusion models generate high-fidelity images, they offer limited control over the spatial and geometric structure of the output. To address this, we introduce and evaluate two ControlNets specialized for artistic control: (1) a proportion ControlNet that uses bounding boxes to dictate the position and scale of objects, and (2) a perspective ControlNet that employs vanishing lines to control the 3D geometry of the scene. We support the training of these modules with data pipelines that leverage vision-language models for annotation and specialized algorithms for conditioning image synthesis. Our experiments demonstrate that both modules provide effective control but exhibit limitations with complex constraints. Both models are released on HuggingFace: https://huggingface.co/obvious-research

---

### A Multi-lingual Dataset of Classified Paragraphs from Open Access Scientific Publications

Authors: Eric Jeangirard
Published: 2025-10-13
Categories: cs.CL, cs.DL
arXiv: https://arxiv.org/abs/2510.21762v1

We present a dataset of 833k paragraphs extracted from CC-BY licensed scientific publications, classified into four categories: acknowledgments, data mentions, software/code mentions, and clinical trial mentions. The paragraphs are primarily in English and French, with additional European languages represented. Each paragraph is annotated with language identification (using fastText) and scientific domain (from OpenAlex). This dataset, derived from the French Open Science Monitor corpus and processed using GROBID, enables training of text classification models and development of named entity recognition systems for scientific literature mining. The dataset is publicly available on HuggingFace https://doi.org/10.57967/hf/6679 under a CC-BY license.

---

### J-ORA: A Framework and Multimodal Dataset for Japanese Object Identification, Reference, Action Prediction in Robot Perception

Authors: Jesse Atuhurra, Hidetaka Kamigaito, Taro Watanabe, Koichiro Yoshino
Published: 2025-10-13
Categories: cs.RO, cs.AI, cs.CV
arXiv: https://arxiv.org/abs/2510.21761v1

We introduce J-ORA, a novel multimodal dataset that bridges the gap in robot perception by providing detailed object attribute annotations within Japanese human-robot dialogue scenarios. J-ORA is designed to support three critical perception tasks, object identification, reference resolution, and next-action prediction, by leveraging a comprehensive template of attributes (e.g., category, color, shape, size, material, and spatial relations). Extensive evaluations with both proprietary and open-source Vision Language Models (VLMs) reveal that incorporating detailed object attributes substantially improves multimodal perception performance compared to without object attributes. Despite the improvement, we find that there still exists a gap between proprietary and open-source VLMs. In addition, our analysis of object affordances demonstrates varying abilities in understanding object functionality and contextual relationships across different VLMs. These findings underscore the importance of rich, context-sensitive attribute annotations in advancing robot perception in dynamic environments. See project page at https://jatuhurrra.github.io/J-ORA/.

---

### Pre to Post-Treatment Glioblastoma MRI Prediction using a Latent Diffusion Model

Authors: Alexandre G. Leclercq, Sébastien Bougleux, Noémie N. Moreau, Alexis Desmonts, Romain Hérault, Aurélien Corroyer-Dulmont
Published: 2025-10-13
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2510.17851v1

Glioblastoma (GBM) is an aggressive primary brain tumor with a median survival of approximately 15 months. In clinical practice, the Stupp protocol serves as the standard first-line treatment. However, patients exhibit highly heterogeneous therapeutic responses which required at least two months before first visual impact can be observed, typically with MRI. Early prediction treatment response is crucial for advancing personalized medicine. Disease Progression Modeling (DPM) aims to capture the trajectory of disease evolution, while Treatment Response Prediction (TRP) focuses on assessing the impact of therapeutic interventions. Whereas most TRP approaches primarly rely on timeseries data, we consider the problem of early visual TRP as a slice-to-slice translation model generating post-treatment MRI from a pre-treatment MRI, thus reflecting the tumor evolution. To address this problem we propose a Latent Diffusion Model with a concatenation-based conditioning from the pre-treatment MRI and the tumor localization, and a classifier-free guidance to enhance generation quality using survival information, in particular post-treatment tumor evolution. Our model were trained and tested on a local dataset consisting of 140 GBM patients collected at Centre François Baclesse. For each patient we collected pre and post T1-Gd MRI, tumor localization manually delineated in the pre-treatment MRI by medical experts, and survival information.

---

### Neural networks for neurocomputing circuits: a computational study of tolerance to noise and activation function non-uniformity when machine learning materials properties

Authors: Ye min Thant, Methawee Nukunudompanich, Chu-Chen Chueh, Manabu Ihara, Sergei Manzhos
Published: 2025-10-13
Categories: cs.NE, cs.LG
arXiv: https://arxiv.org/abs/2510.17849v1

Dedicated analog neurocomputing circuits are promising for high-throughput, low power consumption applications of machine learning (ML) and for applications where implementing a digital computer is unwieldy (remote locations; small, mobile, and autonomous devices, extreme conditions, etc.). Neural networks (NN) implemented in such circuits, however, must contend with circuit noise and the non-uniform shapes of the neuron activation function (NAF) due to the dispersion of performance characteristics of circuit elements (such as transistors or diodes implementing the neurons). We present a computational study of the impact of circuit noise and NAF inhomogeneity in function of NN architecture and training regimes. We focus on one application that requires high-throughput ML: materials informatics, using as representative problem ML of formation energies vs. lowest-energy isomer of peri-condensed hydrocarbons, formation energies and band gaps of double perovskites, and zero point vibrational energies of molecules from QM9 dataset. We show that NNs generally possess low noise tolerance with the model accuracy rapidly degrading with noise level. Single-hidden layer NNs, and NNs with larger-than-optimal sizes are somewhat more noise-tolerant. Models that show less overfitting (not necessarily the lowest test set error) are more noise-tolerant. Importantly, we demonstrate that the effect of activation function inhomogeneity can be palliated by retraining the NN using practically realized shapes of NAFs.

---

### AlgoTrace: Algorithmic Primitives and Compositional Geometry of Reasoning in Language Models

Authors: Samuel Lippl, Thomas McGee, Kimberly Lopez, Ziwen Pan, Pierce Zhang, Salma Ziadi, Oliver Eberle, Ida Momennejad
Published: 2025-10-13
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.15987v3

How do inference time and latent computations enable large language models (LLMs) to solve multi-step reasoning problems? We introduce AlgoTrace, a framework for tracing and steering algorithmic operations in the model latent space for multi-step reasoning. We operationalize primitives by clustering latent activations of the model when solving four benchmarks: Traveling Salesperson Problem (TSP), 3SAT, AIME, and Graph Navigation. We annotate the clusters using their corresponding tokens in the reasoning trace. We then apply function vector methods to extract primitive vectors as reusable compositional building blocks of reasoning. We find that a) injecting a primitive vector into models (Phi, Qwen, Llama) elicits the associated algorithmic operation in the reasoning trace, b) injecting primitives can steer behavior across tasks, c) primitive vectors can be composed through algebraic operations, revealing a geometric logic in activation space, and d) a fine-tuned model exhibits improved composition of primitives (Phi-4-Reasoning vs. Phi-4). These findings demonstrate that LLM reasoning can be understood as a walk through algorithmic primitives in the latent space governed by compositional geometry. These primitives transfer across tasks, and reasoning finetuning strengthens algorithmic generalization and composition across domains.

---

### User Profiles of Sleep Disorder Sufferers: Towards Explainable Clustering and Differential Variable Analysis

Authors: Sifeddine Sellami, Juba Agoun, Lamia Yessad, Louenas Bounia
Published: 2025-10-13
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.15986v1

Sleep disorders have a major impact on patients' health and quality of life, but their diagnosis remains complex due to the diversity of symptoms. Today, technological advances, combined with medical data analysis, are opening new perspectives for a better understanding of these disorders. In particular, explainable artificial intelligence (XAI) aims to make AI model decisions understandable and interpretable for users. In this study, we propose a clustering-based method to group patients according to different sleep disorder profiles. By integrating an explainable approach, we identify the key factors influencing these pathologies. An experiment on anonymized real data illustrates the effectiveness and relevance of our approach.

---

### MEET-Sepsis: Multi-Endogenous-View Enhanced Time-Series Representation Learning for Early Sepsis Prediction

Authors: Zexi Tan, Tao Xie, Binbin Sun, Xiang Zhang, Yiqun Zhang, Yiu-Ming Cheung
Published: 2025-10-13
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.15985v2

Sepsis is a life-threatening infectious syndrome associated with high mortality in intensive care units (ICUs). Early and accurate sepsis prediction (SP) is critical for timely intervention, yet remains challenging due to subtle early manifestations and rapidly escalating mortality. While AI has improved SP efficiency, existing methods struggle to capture weak early temporal signals. This paper introduces a Multi-Endogenous-view Representation Enhancement (MERE) mechanism to construct enriched feature views, coupled with a Cascaded Dual-convolution Time-series Attention (CDTA) module for multi-scale temporal representation learning. The proposed MEET-Sepsis framework achieves competitive prediction accuracy using only 20% of the ICU monitoring time required by SOTA methods, significantly advancing early SP. Extensive validation confirms its efficacy. Code is available at: https://github.com/yueliangy/MEET-Sepsis.

---

### Ontologies in Motion: A BFO-Based Approach to Knowledge Graph Construction for Motor Performance Research Data in Sports Science

Authors: Sarah Rebecca Ondraszek, Jörg Waitelonis, Katja Keller, Claudia Niessner, Anna M. Jacyszyn, Harald Sack
Published: 2025-10-13
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.15983v1

An essential component for evaluating and comparing physical and cognitive capabilities between populations is the testing of various factors related to human performance. As a core part of sports science research, testing motor performance enables the analysis of the physical health of different demographic groups and makes them comparable. The Motor Research (MO|RE) data repository, developed at the Karlsruhe Institute of Technology, is an infrastructure for publishing and archiving research data in sports science, particularly in the field of motor performance research. In this paper, we present our vision for creating a knowledge graph from MO|RE data. With an ontology rooted in the Basic Formal Ontology, our approach centers on formally representing the interrelation of plan specifications, specific processes, and related measurements. Our goal is to transform how motor performance data are modeled and shared across studies, making it standardized and machine-understandable. The idea presented here is developed within the Leibniz Science Campus ``Digital Transformation of Research'' (DiTraRe).

---

### AMiD: Knowledge Distillation for LLMs with $α$-mixture Assistant Distribution

Authors: Donghyeok Shin, Yeongmin Kim, Suhyeon Jo, Byeonghu Na, Il-Chul Moon
Published: 2025-10-13
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.15982v3

Autoregressive large language models (LLMs) have achieved remarkable improvement across many tasks but incur high computational and memory costs. Knowledge distillation (KD) mitigates this issue by transferring knowledge from a large teacher to a smaller student through distributional alignment. Previous studies have proposed various discrepancy metrics, but the capacity gap and training instability caused by near-zero probabilities, stemming from the high-dimensional output of LLMs, remain fundamental limitations. To overcome these challenges, several approaches implicitly or explicitly incorporating assistant distribution have recently been proposed. However, the past proposals of assistant distributions have been a fragmented approach without a systematic investigation of the interpolation path and the divergence. This paper proposes $α$-mixture assistant distribution, a novel generalized family of assistant distributions, and $α$-mixture distillation, coined AMiD, a unified framework for KD using the assistant distribution. The $α$-mixture assistant distribution provides a continuous extension of the assistant distribution by introducing a new distribution design variable $α$, which has been fixed in all previous approaches. Furthermore, AMiD generalizes the family of divergences used with the assistant distributions based on optimality, which has also been restricted in previous works. Through extensive experiments, we demonstrate that AMiD offers superior performance and training stability by leveraging a broader and theoretically grounded assistant distribution space. We release the code at https://github.com/aailab-kaist/AMiD.

---

### ProofFlow: A Dependency Graph Approach to Faithful Proof Autoformalization

Authors: Rafael Cabral, Tuan Manh Do, Xuejun Yu, Wai Ming Tai, Zijin Feng, Xin Shen
Published: 2025-10-13
Categories: cs.AI, cs.LO
arXiv: https://arxiv.org/abs/2510.15981v1

Proof autoformalization, the task of translating natural language theorems and proofs into machine-verifiable code, is a critical step for integrating large language models into rigorous mathematical workflows. Current approaches focus on producing executable code, but they frequently fail to preserve the semantic meaning and logical structure of the original human-written argument. To address this, we introduce ProofFlow, a novel pipeline that treats structural fidelity as a primary objective. ProofFlow first constructs a directed acyclic graph (DAG) to map the logical dependencies between proof steps. Then, it employs a novel lemma-based approach to systematically formalize each step as an intermediate lemma, preserving the logical structure of the original argument. To facilitate evaluation, we present a new benchmark of 184 undergraduate-level problems, manually annotated with step-by-step solutions and logical dependency graphs, and introduce ProofScore, a new composite metric to evaluate syntactic correctness, semantic faithfulness, and structural fidelity. Experimental results show our pipeline sets a new state-of-the-art for autoformalization, achieving a ProofScore of 0.545, substantially exceeding baselines like full-proof formalization (0.123), which processes the entire proof at once, and step-proof formalization (0.072), which handles each step independently. Our pipeline, benchmark, and score metric are open-sourced to encourage further progress at https://github.com/Huawei-AI4Math/ProofFlow.

---

### Cognitive Load Traces as Symbolic and Visual Accounts of Deep Model Cognition

Authors: Dong Liu, Yanxuan Yu
Published: 2025-10-13
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.15980v1

We propose \textbf{Cognitive Load Traces} (CLTs) as a mid-level interpretability framework for deep models, inspired by Cognitive Load Theory in human cognition. CLTs are defined as symbolic, temporally varying functions that quantify model-internal resource allocation. Formally, we represent CLTs as a three-component stochastic process $(\mathrm{IL}_t, \mathrm{EL}_t, \mathrm{GL}_t)$, corresponding to \emph{Intrinsic}, \emph{Extraneous}, and \emph{Germane} load. Each component is instantiated through measurable proxies such as attention entropy, KV-cache miss ratio, representation dispersion, and decoding stability. We propose both symbolic formulations and visualization methods (load curves, simplex diagrams) that enable interpretable analysis of reasoning dynamics. Experiments on reasoning and planning benchmarks show that CLTs predict error-onset, reveal cognitive strategies, and enable load-guided interventions that improve reasoning efficiency by 15-30\% while maintaining accuracy.

---

### Cog-Rethinker: Hierarchical Metacognitive Reinforcement Learning for LLM Reasoning

Authors: Zexu Sun, Yongcheng Zeng, Erxue Min, Heyang Gao, Bokai Ji, Xu Chen
Published: 2025-10-13
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.15979v1

Contemporary progress in large language models (LLMs) has revealed notable inferential capacities via reinforcement learning (RL) employing verifiable reward, facilitating the development of O1 and R1-like reasoning models. Directly training from base models with RL is called zero-RL. However, previous works rely upon activating LLMs' inherent capacities through fixed prompt templates. This strategy introduces substantial sampling inefficiencies for weak LLMs, as the majority of problems generate invalid outputs during accuracy-driven filtration in reasoning tasks, which causes a waste of samples. To solve this issue, we propose Cog-Rethinker, a novel hierarchical metacognitive RL framework for LLM reasoning. Our Cog-Rethinker mainly focuses on the rollout procedure in RL training. After the direct rollout, our Cog-Rethinker improves sample utilization in a hierarchical metacognitive two-stage framework. By leveraging human cognition during solving problems, firstly, it prompts policy to decompose zero-accuracy problems into subproblems to produce final reasoning results. Secondly, with zero-accuracy problems in previous rollout stage, it further prompts policy to refine these answers by referencing previous wrong solutions. Moreover, to enable cold-start of the two new reasoning patterns and maintain train-test consistency across prompt templates, our Cog-Rethinker applies supervised fine-tuning on the policy using correct samples of the two stages with direct rollout template. Experimental results demonstrate Cog-Rethinker's superior performance on various mathematical reasoning benchmarks, we also analyzed its improved sample efficiency that accelerates convergence compared to baseline methods.

---

### DAWP: A framework for global observation forecasting via Data Assimilation and Weather Prediction in satellite observation space

Authors: Junchao Gong, Jingyi Xu, Ben Fei, Fenghua Ling, Wenlong Zhang, Kun Chen, Wanghan Xu, Weidong Yang, Xiaokang Yang, Lei Bai
Published: 2025-10-13
Categories: cs.LG, cs.AI, physics.ao-ph
arXiv: https://arxiv.org/abs/2510.15978v1

Weather prediction is a critical task for human society, where impressive progress has been made by training artificial intelligence weather prediction (AIWP) methods with reanalysis data. However, reliance on reanalysis data limits the AIWPs with shortcomings, including data assimilation biases and temporal discrepancies. To liberate AIWPs from the reanalysis data, observation forecasting emerges as a transformative paradigm for weather prediction. One of the key challenges in observation forecasting is learning spatiotemporal dynamics across disparate measurement systems with irregular high-resolution observation data, which constrains the design and prediction of AIWPs. To this end, we propose our DAWP as an innovative framework to enable AIWPs to operate in a complete observation space by initialization with an artificial intelligence data assimilation (AIDA) module. Specifically, our AIDA module applies a mask multi-modality autoencoder(MMAE)for assimilating irregular satellite observation tokens encoded by mask ViT-VAEs. For AIWP, we introduce a spatiotemporal decoupling transformer with cross-regional boundary conditioning (CBC), learning the dynamics in observation space, to enable sub-image-based global observation forecasting. Comprehensive experiments demonstrate that AIDA initialization significantly improves the roll out and efficiency of AIWP. Additionally, we show that DAWP holds promising potential to be applied in global precipitation forecasting.

---

### Bolster Hallucination Detection via Prompt-Guided Data Augmentation

Authors: Wenyun Li, Zheng Zhang, Dongmei Jiang, Xiangyuan Lan
Published: 2025-10-13
Categories: cs.LG, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2510.15977v1

Large language models (LLMs) have garnered significant interest in AI community. Despite their impressive generation capabilities, they have been found to produce misleading or fabricated information, a phenomenon known as hallucinations. Consequently, hallucination detection has become critical to ensure the reliability of LLM-generated content. One primary challenge in hallucination detection is the scarcity of well-labeled datasets containing both truthful and hallucinated outputs. To address this issue, we introduce Prompt-guided data Augmented haLlucination dEtection (PALE), a novel framework that leverages prompt-guided responses from LLMs as data augmentation for hallucination detection. This strategy can generate both truthful and hallucinated data under prompt guidance at a relatively low cost. To more effectively evaluate the truthfulness of the sparse intermediate embeddings produced by LLMs, we introduce an estimation metric called the Contrastive Mahalanobis Score (CM Score). This score is based on modeling the distributions of truthful and hallucinated data in the activation space. CM Score employs a matrix decomposition approach to more accurately capture the underlying structure of these distributions. Importantly, our framework does not require additional human annotations, offering strong generalizability and practicality for real-world applications. Extensive experiments demonstrate that PALE achieves superior hallucination detection performance, outperforming the competitive baseline by a significant margin of 6.55%.

---

### Learning to Watermark: A Selective Watermarking Framework for Large Language Models via Multi-Objective Optimization

Authors: Chenrui Wang, Junyi Shu, Billy Chiu, Yu Li, Saleh Alharbi, Min Zhang, Jing Li
Published: 2025-10-13
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2510.15976v1

The rapid development of LLMs has raised concerns about their potential misuse, leading to various watermarking schemes that typically offer high detectability. However, existing watermarking techniques often face trade-off between watermark detectability and generated text quality. In this paper, we introduce Learning to Watermark (LTW), a novel selective watermarking framework that leverages multi-objective optimization to effectively balance these competing goals. LTW features a lightweight network that adaptively decides when to apply the watermark by analyzing sentence embeddings, token entropy, and current watermarking ratio. Training of the network involves two specifically constructed loss functions that guide the model toward Pareto-optimal solutions, thereby harmonizing watermark detectability and text quality. By integrating LTW with two baseline watermarking methods, our experimental evaluations demonstrate that LTW significantly enhances text quality without compromising detectability. Our selective watermarking approach offers a new perspective for designing watermarks for LLMs and a way to preserve high text quality for watermarks. The code is publicly available at: https://github.com/fattyray/learning-to-watermark

---

### PAGE: Prompt Augmentation for text Generation Enhancement

Authors: Mauro Jose Pacchiotti, Luciana Ballejos, Mariel Ale
Published: 2025-10-13
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.13880v1

In recent years, natural language generative models have shown outstanding performance in text generation tasks. However, when facing specific tasks or particular requirements, they may exhibit poor performance or require adjustments that demand large amounts of additional data. This work introduces PAGE (Prompt Augmentation for text Generation Enhancement), a framework designed to assist these models through the use of simple auxiliary modules. These modules, lightweight models such as classifiers or extractors, provide inferences from the input text. The output of these auxiliaries is then used to construct an enriched input that improves the quality and controllability of the generation. Unlike other generation-assistance approaches, PAGE does not require auxiliary generative models; instead, it proposes a simpler, modular architecture that is easy to adapt to different tasks. This paper presents the proposal, its components and architecture, and reports a proof of concept in the domain of requirements engineering, where an auxiliary module with a classifier is used to improve the quality of software requirements generation.

---

### Catch Your Breath: Adaptive Computation for Self-Paced Sequence Production

Authors: Alexandre Galashov, Matt Jones, Rosemary Ke, Yuan Cao, Vaishnavh Nagarajan, Michael C. Mozer
Published: 2025-10-13
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.13879v2

Within the landscape of inference-time scaling methods for foundation models, a width-based approach to scaling -- which involves the insertion of <pause> tokens in the input stream to delay model responses -- offers a unique advantage by increasing model expressivity while remaining highly parallelizable at both training and inference. The existing literature on training models to utilize <pause> tokens relies on the standard cross-entropy objective in which the model output is read out and evaluated only at the final step of a pause sequence. This approach provides no mechanism for the model to regulate its own processing or to signal readiness to respond, treating the additional compute steps as a static barrier rather than a resource to be used adaptively. We propose a supervised loss, Catch Your Breath (CYB), framed as a sequential-decision problem, that trains a model to dynamically and autonomously scale the number of compute steps used for each input token. The model indicates the need for additional compute steps by emitting a special <don't know> output, delaying its response via a pause. The model can abstain multiple times to obtain longer delays. Our experiments demonstrate that CYB significantly outperforms standard cross-entropy when introduced either in pretraining or fine-tuning, reducing perplexity and enhancing downstream accuracy with no additional computational or memory cost.

---

### TextBandit: Evaluating Probabilistic Reasoning in LLMs Through Language-Only Decision Tasks

Authors: Jimin Lim, Arjun Damerla, Arthur Jiang, Nam Le
Published: 2025-10-13
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.13878v1

Large language models (LLMs) have shown to be increasingly capable of performing reasoning tasks, but their ability to make sequential decisions under uncertainty only using natural language remains underexplored. We introduce a novel benchmark in which LLMs interact with multi-armed bandit environments using purely textual feedback, "you earned a token", without access to numerical cues or explicit probabilities, resulting in the model to infer latent reward structures purely off linguistic cues and to adapt accordingly. We evaluated the performance of four open-source LLMs and compare their performance to standard decision-making algorithms such as Thompson Sampling, Epsilon Greedy, Upper Confidence Bound (UCB), and random choice. While most of the LLMs underperformed compared to the baselines, Qwen3-4B, achieved the best-arm selection rate of 89.2% , which significantly outperformed both the larger LLMs and traditional methods. Our findings suggest that probabilistic reasoning is able to emerge from language alone, and we present this benchmark as a step towards evaluating decision-making capabilities in naturalistic, non-numeric contexts.

---

### What Layers When: Learning to Skip Compute in LLMs with Residual Gates

Authors: Filipe Laitenberger, Dawid Kopiczko, Cees G. M. Snoek, Yuki M. Asano
Published: 2025-10-13
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.13876v3

We introduce GateSkip, a simple residual-stream gating mechanism that enables token-wise layer skipping in decoder-only LMs. Each Attention/MLP branch is equipped with a sigmoid-linear gate that condenses the branch's output before it re-enters the residual stream. During inference we rank tokens by the gate values and skip low-importance ones using a per-layer budget. While early-exit or router-based Mixture-of-Depths models are known to be unstable and need extensive retraining, our smooth, differentiable gates fine-tune stably on top of pretrained models. On long-form reasoning, we save up to 15% compute while retaining over 90% of baseline accuracy. For increasingly larger models, this tradeoff improves drastically. On instruction-tuned models we see accuracy gains at full compute and match baseline quality near 50% savings. The learned gates give insight into transformer information flow (e.g., BOS tokens act as anchors), and the method combines easily with quantization, pruning, and self-speculative decoding.

---

### FRACCO: A gold-standard annotated corpus of oncological entities with ICD-O-3.1 normalisation

Authors: Johann Pignat, Milena Vucetic, Christophe Gaudet-Blavignac, Jamil Zaghir, Amandine Stettler, Fanny Amrein, Jonatan Bonjour, Jean-Philippe Goldman, Olivier Michielin, Christian Lovis, Mina Bjelogrlic
Published: 2025-10-13
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.13873v1

Developing natural language processing tools for clinical text requires annotated datasets, yet French oncology resources remain scarce. We present FRACCO (FRench Annotated Corpus for Clinical Oncology) an expert-annotated corpus of 1301 synthetic French clinical cases, initially translated from the Spanish CANTEMIST corpus as part of the FRASIMED initiative. Each document is annotated with terms related to morphology, topography, and histologic differentiation, using the International Classification of Diseases for Oncology (ICD-O) as reference. An additional annotation layer captures composite expression-level normalisations that combine multiple ICD-O elements into unified clinical concepts. Annotation quality was ensured through expert review: 1301 texts were manually annotated for entity spans by two domain experts. A total of 71127 ICD-O normalisations were produced through a combination of automated matching and manual validation by a team of five annotators. The final dataset representing 399 unique morphology codes (from 2549 different expressions), 272 topography codes (from 3143 different expressions), and 2043 unique composite expressions (from 11144 different expressions). This dataset provides a reference standard for named entity recognition and concept normalisation in French oncology texts.

---

### Scalable Energy-Based Models via Adversarial Training: Unifying Discrimination and Generation

Authors: Xuwang Yin, Claire Zhang, Julie Steele, Nir Shavit, Tony T. Wang
Published: 2025-10-13
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.13872v4

Simultaneously achieving robust classification and high-fidelity generative modeling within a single framework presents a significant challenge. Hybrid approaches, such as Joint Energy-Based Models (JEM), interpret classifiers as EBMs but are often limited by the instability and poor sample quality inherent in training based on Stochastic Gradient Langevin Dynamics (SGLD). We address these limitations by proposing a novel training framework that integrates adversarial training (AT) principles for both discriminative robustness and stable generative learning. The proposed method introduces three key innovations: (1) the replacement of SGLD-based JEM learning with a stable, AT-based approach that optimizes the energy function through a Binary Cross-Entropy (BCE) loss that discriminates between real data and contrastive samples generated via Projected Gradient Descent (PGD); (2) adversarial training for the discriminative component that enhances classification robustness while implicitly providing the gradient regularization needed for stable EBM training; and (3) a two-stage training strategy that addresses normalization-related instabilities and enables leveraging pretrained robust classifiers, generalizing effectively across architectures. Experiments on CIFAR-10/100 and ImageNet demonstrate that our approach: (1) is the first EBM-based hybrid to scale to high-resolution datasets with high training stability, simultaneously achieving state-of-the-art discriminative and generative performance on ImageNet 256x256; (2) uniquely combines generative quality with adversarial robustness, enabling faithful counterfactual explanations; and (3) functions as a competitive standalone generative model, matching autoregressive models and surpassing diffusion models while offering additional versatility.

---

### Quechua Speech Datasets in Common Voice: The Case of Puno Quechua

Authors: Elwin Huaman, Wendi Huaman, Jorge Luis Huaman, Ninfa Quispe
Published: 2025-10-13
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.13871v1

Under-resourced languages, such as Quechuas, face data and resource scarcity, hindering their development in speech technology. To address this issue, Common Voice presents a crucial opportunity to foster an open and community-driven speech dataset creation. This paper examines the integration of Quechua languages into Common Voice. We detail the current 17 Quechua languages, presenting Puno Quechua (ISO 639-3: qxp) as a focused case study that includes language onboarding and corpus collection of both reading and spontaneous speech data. Our results demonstrate that Common Voice now hosts 191.1 hours of Quechua speech (86\% validated), with Puno Quechua contributing 12 hours (77\% validated), highlighting the Common Voice's potential. We further propose a research agenda addressing technical challenges, alongside ethical considerations for community engagement and indigenous data sovereignty. Our work contributes towards inclusive voice technology and digital empowerment of under-resourced language communities.

---

### Unlocking the Potential of Diffusion Language Models through Template Infilling

Authors: Junhoo Lee, Seungyeon Kim, Nojun Kwak
Published: 2025-10-13
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.13870v3

Diffusion Language Models (DLMs) have emerged as a promising alternative to Autoregressive Language Models, yet their inference strategies remain limited to prefix-based prompting inherited from the autoregressive paradigm. In this paper, we propose Template Infilling (TI), a tailored conditioning methodology for DLMs. Unlike conventional prefix prompting, TI flexibly aligns structural anchors across the entire target response space, establishing a global blueprint before filling in the masked segments. We demonstrate the effectiveness of our approach on diverse benchmarks, including mathematical reasoning, code generation, and trip planning, achieving consistent improvements of 9.40% over the baseline. Furthermore, we observe that TI provides additional advantages in multi-token generation settings, enabling effective speedup while maintaining generation quality and robustness. By enforcing these global constraints, TI ultimately facilitates System-2 reasoning, empowering the model to deliberate within a structurally defined solution space.

---

### CoLoR-GAN: Continual Few-Shot Learning with Low-Rank Adaptation in Generative Adversarial Networks

Authors: Munsif Ali, Leonardo Rossi, Massimo Bertozzi
Published: 2025-10-13
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.13869v1

Continual learning (CL) in the context of Generative Adversarial Networks (GANs) remains a challenging problem, particularly when it comes to learn from a few-shot (FS) samples without catastrophic forgetting. Current most effective state-of-the-art (SOTA) methods, like LFS-GAN, introduce a non-negligible quantity of new weights at each training iteration, which would become significant when considering the long term. For this reason, this paper introduces \textcolor{red}{\textbf{\underline{c}}}ontinual few-sh\textcolor{red}{\textbf{\underline{o}}}t learning with \textcolor{red}{\textbf{\underline{lo}}}w-\textcolor{red}{\textbf{\underline{r}}}ank adaptation in GANs named CoLoR-GAN, a framework designed to handle both FS and CL together, leveraging low-rank tensors to efficiently adapt the model to target tasks while reducing even more the number of parameters required. Applying a vanilla LoRA implementation already permitted us to obtain pretty good results. In order to optimize even further the size of the adapters, we challenged LoRA limits introducing a LoRA in LoRA (LLoRA) technique for convolutional layers. Finally, aware of the criticality linked to the choice of the hyperparameters of LoRA, we provide an empirical study to easily find the best ones. We demonstrate the effectiveness of CoLoR-GAN through experiments on several benchmark CL and FS tasks and show that our model is efficient, reaching SOTA performance but with a number of resources enormously reduced. Source code is available on \href{https://github.com/munsifali11/CoLoR-GAN}{Github.

---

### DeepMartingale: Duality of the Optimal Stopping Problem with Expressivity and High-Dimensional Hedging

Authors: Junyan Ye, Hoi Ying Wong
Published: 2025-10-13
Categories: math.OC, cs.LG, math.NA, math.PR, stat.ML
arXiv: https://arxiv.org/abs/2510.13868v2

We propose \textit{DeepMartingale}, a deep-learning framework for the dual formulation of discrete-monitoring optimal stopping problems under continuous-time models. Leveraging a martingale representation, our method implements a \emph{pure-dual} procedure that directly optimizes over a parameterized class of martingales, producing computable and tight \emph{dual upper bounds} for the value function in high-dimensional settings without requiring any primal information or Snell-envelope approximation. We prove convergence of the resulting upper bounds under mild assumptions for both first- and second-moment losses. A key contribution is an expressivity theorem showing that \textit{DeepMartingale} can approximate the true value function to any prescribed accuracy $\varepsilon$ using neural networks of size at most $\tilde{c} d^{\tilde{q}}\varepsilon^{-\tilde{r}}$, with constants independent of the dimension $d$ and accuracy $\varepsilon$, thereby avoiding the curse of dimensionality. Since expressivity in this setting translates into scalability, our theory also motivates estimating the dimension scaling law to guide architecture design and the training setup in deep learning-based numerical computation and the choice of rebalancing frequency for the related hedging strategy. The learned martingale representation further yields a practical and dimension-scalable \emph{deep delta hedging strategy}. Numerical experiments on high-dimensional Bermudan option benchmarks confirm convergence, expressivity, scalable training, and the stability of the resulting upper bounds and hedging performance.

---

### An Overview of the JPEG AI Learning-Based Image Coding Standard

Authors: Semih Esenlik, Yaojun Wu, Zhaobin Zhang, Ye-Kui Wang, Kai Zhang, Li Zhang, João Ascenso, Shan Liu
Published: 2025-10-13
Categories: eess.IV, cs.LG, cs.MM
arXiv: https://arxiv.org/abs/2510.13867v1

JPEG AI is an emerging learning-based image coding standard developed by Joint Photographic Experts Group (JPEG). The scope of the JPEG AI is the creation of a practical learning-based image coding standard offering a single-stream, compact compressed domain representation, targeting both human visualization and machine consumption. Scheduled for completion in early 2025, the first version of JPEG AI focuses on human vision tasks, demonstrating significant BD-rate reductions compared to existing standards, in terms of MS-SSIM, FSIM, VIF, VMAF, PSNR-HVS, IW-SSIM and NLPD quality metrics. Designed to ensure broad interoperability, JPEG AI incorporates various design features to support deployment across diverse devices and applications. This paper provides an overview of the technical features and characteristics of the JPEG AI standard.

---

### FFT-Accelerated Auxiliary Variable MCMC for Fermionic Lattice Models: A Determinant-Free Approach with $O(N\log N)$ Complexity

Authors: Deqian Kong, Shi Feng, Jianwen Xie, Ying Nian Wu
Published: 2025-10-13
Categories: cond-mat.str-el, cs.AI, cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2510.13866v1

We introduce a Markov Chain Monte Carlo (MCMC) algorithm that dramatically accelerates the simulation of quantum many-body systems, a grand challenge in computational science. State-of-the-art methods for these problems are severely limited by $O(N^3)$ computational complexity. Our method avoids this bottleneck, achieving near-linear $O(N \log N)$ scaling per sweep. Our approach samples a joint probability measure over two coupled variable sets: (1) particle trajectories of the fundamental fermions, and (2) auxiliary variables that decouple fermion interactions. The key innovation is a novel transition kernel for particle trajectories formulated in the Fourier domain, revealing the transition probability as a convolution that enables massive acceleration via the Fast Fourier Transform (FFT). The auxiliary variables admit closed-form, factorized conditional distributions, enabling efficient exact Gibbs sampling update. We validate our algorithm on benchmark quantum physics problems, accurately reproducing known theoretical results and matching traditional $O(N^3)$ algorithms on $32\times 32$ lattice simulations at a fraction of the wall-clock time, empirically demonstrating $N \log N$ scaling. By reformulating a long-standing physics simulation problem in machine learning language, our work provides a powerful tool for large-scale probabilistic inference and opens avenues for physics-inspired generative models.

---

### Deep Edge Filter: Return of the Human-Crafted Layer in Deep Learning

Authors: Dongkwan Lee, Junhoo Lee, Nojun Kwak
Published: 2025-10-13
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2510.13865v5

We introduce the Deep Edge Filter, a novel approach that applies high-pass filtering to deep neural network features to improve model generalizability. Our method is motivated by our hypothesis that neural networks encode task-relevant semantic information in high-frequency components while storing domain-specific biases in low-frequency components of deep features. By subtracting low-pass filtered outputs from original features, our approach isolates generalizable representations while preserving architectural integrity. Experimental results across diverse domains such as Vision, Text, 3D, and Audio demonstrate consistent performance improvements regardless of model architecture and data modality. Analysis reveals that our method induces feature sparsification and effectively isolates high-frequency components, providing empirical validation of our core hypothesis. The code is available at https://github.com/dongkwani/DeepEdgeFilter.

---

### Self-Training with Dynamic Weighting for Robust Gradual Domain Adaptation

Authors: Zixi Wang, Yushe Cao, Yubo Huang, Jinzhu Wei, Jingzehua Xu, Shuai Zhang, Xin Lai
Published: 2025-10-13
Categories: cs.LG, cs.AI, cs.CV
arXiv: https://arxiv.org/abs/2510.13864v1

In this paper, we propose a new method called Self-Training with Dynamic Weighting (STDW), which aims to enhance robustness in Gradual Domain Adaptation (GDA) by addressing the challenge of smooth knowledge migration from the source to the target domain. Traditional GDA methods mitigate domain shift through intermediate domains and self-training but often suffer from inefficient knowledge migration or incomplete intermediate data. Our approach introduces a dynamic weighting mechanism that adaptively balances the loss contributions of the source and target domains during training. Specifically, we design an optimization framework governed by a time-varying hyperparameter $\varrho$ (progressing from 0 to 1), which controls the strength of domain-specific learning and ensures stable adaptation. The method leverages self-training to generate pseudo-labels and optimizes a weighted objective function for iterative model updates, maintaining robustness across intermediate domains. Experiments on rotated MNIST, color-shifted MNIST, portrait datasets, and the Cover Type dataset demonstrate that STDW outperforms existing baselines. Ablation studies further validate the critical role of $\varrho$'s dynamic scheduling in achieving progressive adaptation, confirming its effectiveness in reducing domain bias and improving generalization. This work provides both theoretical insights and a practical framework for robust gradual domain adaptation, with potential applications in dynamic real-world scenarios. The code is available at https://github.com/Dramwig/STDW.

---

### Ensembling Large Language Models to Characterize Affective Dynamics in Student-AI Tutor Dialogues

Authors: Chenyu Zhang, Sharifa Alghowinem, Cynthia Breazeal
Published: 2025-10-13
Categories: cs.CL, cs.AI, cs.HC
arXiv: https://arxiv.org/abs/2510.13862v1

While recent studies have examined the leaning impact of large language model (LLM) in educational contexts, the affective dynamics of LLM-mediated tutoring remain insufficiently understood. This work introduces the first ensemble-LLM framework for large-scale affect sensing in tutoring dialogues, advancing the conversation on responsible pathways for integrating generative AI into education by attending to learners' evolving affective states. To achieve this, we analyzed two semesters' worth of 16,986 conversational turns exchanged between PyTutor, an LLM-powered AI tutor, and 261 undergraduate learners across three U.S. institutions. To investigate learners' emotional experiences, we generate zero-shot affect annotations from three frontier LLMs (Gemini, GPT-4o, Claude), including scalar ratings of valence, arousal, and learning-helpfulness, along with free-text emotion labels. These estimates are fused through rank-weighted intra-model pooling and plurality consensus across models to produce robust emotion profiles. Our analysis shows that during interaction with the AI tutor, students typically report mildly positive affect and moderate arousal. Yet learning is not uniformly smooth: confusion and curiosity are frequent companions to problem solving, and frustration, while less common, still surfaces in ways that can derail progress. Emotional states are short-lived--positive moments last slightly longer than neutral or negative ones, but they are fragile and easily disrupted. Encouragingly, negative emotions often resolve quickly, sometimes rebounding directly into positive states. Neutral moments frequently act as turning points, more often steering students upward than downward, suggesting opportunities for tutors to intervene at precisely these junctures.

---

### ShishuLM : Achieving Optimal and Efficient Parameterization with Low Attention Transformer Models

Authors: Shivanshu Kumar, Gopalakrishnan Srinivasan
Published: 2025-10-13
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.13860v2

While the transformer architecture has achieved state-of-the-art performance on natural language processing tasks, these models impose substantial memory and computational overhead. Recent research has identified significant architectural redundancies within these models, particularly in the attention sub-layers in the top layers, presenting opportunities for optimization without compromising performance. Taking insights from research on inference-time layer pruning and depth-dependent computation in language models, we introduce an efficient language model architecture referred to as ShishuLM. By replacing full decoder layers at the top of the model with MLP-only blocks, we achieve up to 10-60% improvement in generation latency and 1.3 -5 $\times$ gain in throughput. Upon further sharing parameters across adjacent MLP-only layers of ShishuLM, we obtain up to 20% savings in memory with minimal degradation in performance. Our findings provide insights towards building more efficient language modeling architectures from a pre-training standpoint by leveraging how information flows in transformers.

---

### Benchmarking Correctness and Security in Multi-Turn Code Generation

Authors: Ruchit Rawal, Jeffrey Yang Fan Chiang, Chihao Shen, Jeffery Siyuan Tian, Aastha Mahajan, Tom Goldstein, Yizheng Chen
Published: 2025-10-13
Categories: cs.SE, cs.AI
arXiv: https://arxiv.org/abs/2510.13859v1

AI coding assistants powered by large language models (LLMs) have transformed software development, significantly boosting productivity. While existing benchmarks evaluate the correctness and security of LLM-generated code, they are typically limited to single-turn tasks that do not reflect the iterative nature of real-world development. We introduce MT-Sec, the first benchmark to systematically evaluate both correctness and security in multi-turn coding scenarios. We construct this using a synthetic data pipeline that transforms existing single-turn tasks into semantically aligned multi-turn interaction sequences, allowing reuse of original test suites while modeling the complexity of real-world coding processes. We evaluate 32 open- and closed-source models, and three agent-scaffolding on MT-Sec and observe a consistent 20-27% drop in "correct and secure" outputs from single-turn to multi-turn settings -- even among state-of-the-art models. Beyond full-program generation, we also evaluate models on multi-turn code-diff generation -- an unexplored yet practically relevant setting -- and find that models perform worse here, with increased rates of functionally incorrect and insecure outputs. Finally, we find that while agent scaffoldings boost single-turn code generation performance, they are not quite as effective in multi-turn evaluations. Together, these findings highlight the need for benchmarks that jointly evaluate correctness and security in multi-turn, real-world coding workflows.

---

### Local Timescale Gates for Timescale-Robust Continual Spiking Neural Networks

Authors: Devansh Deep, Ansh Tiwari, Ayush Chauhan
Published: 2025-10-13
Categories: cs.LG
arXiv: https://arxiv.org/abs/2510.12843v2

Spiking neural networks (SNNs) promise energy-efficient artificial intelligence on neuromorphic hardware but struggle with tasks requiring both fast adaptation and long-term memory, especially in continual learning. We propose Local Timescale Gating (LT-Gate), a neuron model that combines dual time-constant dynamics with an adaptive gating mechanism. Each spiking neuron tracks information on a fast and a slow timescale in parallel, and a learned gate locally adjusts their influence. This design enables individual neurons to preserve slow contextual information while responding to fast signals, addressing the stability-plasticity dilemma. We further introduce a variance-tracking regularization that stabilizes firing activity, inspired by biological homeostasis. Empirically, LT-Gate yields significantly improved accuracy and retention in sequential learning tasks: on a challenging temporal classification benchmark it achieves about 51 percent final accuracy, compared to about 46 percent for a recent Hebbian continual-learning baseline and lower for prior SNN methods. Unlike approaches that require external replay or expensive orthogonalizations, LT-Gate operates with local updates and is fully compatible with neuromorphic hardware. In particular, it leverages features of Intel's Loihi chip (multiple synaptic traces with different decay rates) for on-chip learning. Our results demonstrate that multi-timescale gating can substantially enhance continual learning in SNNs, narrowing the gap between spiking and conventional deep networks on lifelong-learning tasks.

---

### Protenix-Mini+: efficient structure prediction model with scalable pairformer

Authors: Bo Qiang, Chengyue Gong, Xinshi Chen, Yuxuan Zhang, Wenzhi Xiao
Published: 2025-10-13
Categories: q-bio.QM, cs.LG
arXiv: https://arxiv.org/abs/2510.12842v2

Lightweight inference is critical for biomolecular structure prediction and downstream tasks, enabling efficient real-world deployment and inference-time scaling for large-scale applications. While AF3 and its variants (e.g., Protenix, Chai-1) have advanced structure prediction results, they suffer from critical limitations: high inference latency and cubic time complexity with respect to token count, both of which restrict scalability for large biomolecular complexes. To address the core challenge of balancing model efficiency and prediction accuracy, we introduce three key innovations: (1) compressing non-scalable operations to mitigate cubic time complexity, (2) removing redundant blocks across modules to reduce unnecessary overhead, and (3) adopting a few-step sampler for the atom diffusion module to accelerate inference. Building on these design principles, we develop Protenix-Mini+, a highly lightweight and scalable variant of the Protenix model. Within an acceptable range of performance degradation, it substantially improves computational efficiency. For example, in the case of low-homology single-chain proteins, Protenix-Mini+ experiences an intra-protein LDDT drop of approximately 3% relative to the full Protenix model -- an acceptable performance trade-off given its substantially 90%+ improved computational efficiency.

---

### FaStfact: Faster, Stronger Long-Form Factuality Evaluations in LLMs

Authors: Yingjia Wan, Haochen Tan, Xiao Zhu, Xinyu Zhou, Zhiwei Li, Qingsong Lv, Changxuan Sun, Jiaqi Zeng, Yi Xu, Jianqiao Lu, Yinhong Liu, Zhijiang Guo
Published: 2025-10-13
Categories: cs.CL, cs.AI, cs.CE, cs.CY
arXiv: https://arxiv.org/abs/2510.12839v2

Evaluating the factuality of long-form generations from Large Language Models (LLMs) remains challenging due to efficiency bottlenecks and reliability concerns. Prior efforts attempt this by decomposing text into claims, searching for evidence, and verifying claims, but suffer from critical drawbacks: (1) inefficiency due to overcomplicated pipeline components, and (2) ineffectiveness stemming from inaccurate claim sets and insufficient evidence. To address these limitations, we propose \textbf{FaStfact}, an evaluation framework that achieves the highest alignment with human evaluation and time/token efficiency among existing baselines. FaStfact first employs chunk-level claim extraction integrated with confidence-based pre-verification, significantly reducing the time and token cost while ensuring reliability. For searching and verification, it collects document-level evidence from crawled web-pages and selectively retrieves it during verification. Extensive experiments based on an annotated benchmark \textbf{FaStfact-Bench} demonstrate the reliability of FaStfact in both efficiently and effectively evaluating long-form factuality. Code, benchmark data, and annotation interface tool are available at https://github.com/Yingjia-Wan/FaStfact.

---

### A$^2$FM: An Adaptive Agent Foundation Model for Tool-Aware Hybrid Reasoning

Authors: Qianben Chen, Jingyi Cao, Jiayu Zhang, Tianrui Qin, Xiaowan Li, King Zhu, Dingfeng Shi, He Zhu, Minghao Liu, Xiaobo Liang, Xin Gui, Ge Zhang et al.
Published: 2025-10-13
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.12838v3

Large language models split into two families: reasoning-centric LLMs, which strengthen internal chain-of-thought reasoning but cannot invoke external tools, and agentic LLMs, which learn to interact with environments and leverage tools but often lag in deep reasoning. This divide arises from fundamentally different training objectives, leading to mismatched strengths and inefficiency on simple queries, where both families tend to overthink or over-call tools. In this work, we present Adaptive Agent Foundation Model (A$^2$FM), a unified framework that follows a route-then-align principle: the model first learns task-aware routing and then aligns mode-specific trajectories under a shared backbone. To address the inefficiency gap, we introduce a third mode-instant-that handles simple queries directly, preventing unnecessary reasoning or tool calls while complementing the agentic and reasoning modes. To jointly enhance accuracy and efficiency, we propose Adaptive Policy Optimization (APO), which enforces adaptive sampling across modes and applies a cost-regularized reward. On the 32B scale, A$^2$FM achieves 13.4% on BrowseComp, 70.4% on AIME25, and 16.7% on HLE, setting new SOTA among comparable models and performing competitively with frontier LLMs across agentic, reasoning, and general benchmarks. Notably, the adaptive execution achieves a cost of pass of only $0.00487 per correct answer-cutting cost by 45.2% relative to reasoning and 33.5% relative to agentic, thus delivering substantially higher cost efficiency while maintaining comparable accuracy.

---

### Semantic knowledge guides innovation and drives cultural evolution

Authors: Anil Yaman, Shen Tian, Björn Lindström
Published: 2025-10-13
Categories: cs.MA, cs.AI, cs.CY, cs.NE
arXiv: https://arxiv.org/abs/2510.12837v4

Cultural evolution allows ideas and technologies to accumulate across generations, reaching their most complex and open-ended form in humans. While social learning enables the transmission of such innovations, the cognitive processes that generate them remain poorly understood. Classical theories typically treat innovation as random variation, a simplification insufficient for explaining the complexity of human cultural evolution. We propose that semantic knowledge-the associations linking concepts to their properties and functions-guides human innovation and drives cumulative culture. To test this, we combined an agent-based model, which examines how semantic knowledge shapes cultural evolutionary dynamics, with a large-scale behavioral experiment (N = 1,243) testing its role in human innovation. Across both approaches, we found that semantic knowledge directed exploration toward meaningful solutions, enhanced innovation success, and enabled generalization from prior discoveries. Moreover, semantic knowledge interacted synergistically with social learning to amplify innovation and accelerate cumulative cultural change. In contrast, experimental participants lacking access to semantic knowledge performed no better than chance, even when social learning was possible, and relied on shallow exploration strategies for innovation. Together, these findings suggest that semantic knowledge is a key cognitive process underpinning human cumulative culture.

---

### Repurposing Annotation Guidelines to Instruct LLM Annotators: A Case Study

Authors: Kon Woo Kim, Rezarta Islamaj, Jin-Dong Kim, Florian Boudin, Akiko Aizawa
Published: 2025-10-13
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2510.12835v1

This study investigates how existing annotation guidelines can be repurposed to instruct large language model (LLM) annotators for text annotation tasks. Traditional guidelines are written for human annotators who internalize training, while LLMs require explicit, structured instructions. We propose a moderation-oriented guideline repurposing method that transforms guidelines into clear directives for LLMs through an LLM moderation process. Using the NCBI Disease Corpus as a case study, our experiments show that repurposed guidelines can effectively guide LLM annotators, while revealing several practical challenges. The results highlight the potential of this workflow to support scalable and cost-effective refinement of annotation guidelines and automated annotation.

---

### Gelina: Unified Speech and Gesture Synthesis via Interleaved Token Prediction

Authors: Téo Guichoux, Théodor Lemerle, Shivam Mehta, Jonas Beskow, Gustav Eje Henter, Laure Soulier, Catherine Pelachaud, Nicolas Obin
Published: 2025-10-13
Categories: cs.SD, cs.AI, eess.AS
arXiv: https://arxiv.org/abs/2510.12834v4

Human communication is multimodal, with speech and gestures tightly coupled, yet most computational methods for generating speech and gestures synthesize them sequentially, weakening synchrony and prosody alignment. We introduce Gelina, a unified framework that jointly synthesizes speech and co-speech gestures from text using interleaved token sequences in a discrete autoregressive backbone, with modality-specific decoders. Gelina supports multi-speaker and multi-style cloning and enables gesture-only synthesis from speech inputs. Subjective and objective evaluations demonstrate competitive speech quality and improved gesture generation over unimodal baselines.

---

### Coherent Load Profile Synthesis with Conditional Diffusion for LV Distribution Network Scenario Generation

Authors: Alistair Brash, Junyi Lu, Bruce Stephen, Blair Brown, Robert Atkinson, Craig Michie, Fraser MacIntyre, Christos Tachtatzis
Published: 2025-10-13
Categories: eess.SY, cs.AI, cs.LG, eess.SP
arXiv: https://arxiv.org/abs/2510.12832v2

Limited visibility of distribution network power flows at the low voltage level presents challenges to both distribution network operators from a planning perspective and distribution system operators from a congestion management perspective. More representative loads are required to support meaningful analysis of LV substations; otherwise, such analysis risks misinforming future decisions. Traditional load profiling relies on typical profiles, oversimplifying substation-level complexity. Generative models have attempted to address this through synthesising representative loads from historical exemplars; however, while these approaches can approximate load shapes to a convincing degree of fidelity, analysis of the co-behaviour between substations is limited, which ultimately impacts higher voltage level network operation. This limitation will become even more pronounced with the increasing integration of low-carbon technologies, as estimates of base loads fail to capture load diversity. To address this gap, Conditional Diffusion models for synthesising daily active and reactive power profiles at the low voltage distribution substation level are proposed. The evaluation of fidelity is demonstrated through conventional metrics capturing temporal and statistical realism, as well as power flow modelling. Multiple models are proposed to handle varying levels of data availability, ranging from unconditional synthesis to an informed generation driven by metadata and daily statistics. The results show synthesised load profiles are plausible both independently and as a cohort in a wider power systems context. The Conditional Diffusion model is benchmarked against both naive and state-of-the-art models to demonstrate its effectiveness in producing realistic scenarios on which to base sub-regional power distribution network planning and operations.

---

### Evaluating the Explainability of Vision Transformers in Medical Imaging

Authors: Leili Barekatain, Ben Glocker
Published: 2025-10-13
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.12021v1

Understanding model decisions is crucial in medical imaging, where interpretability directly impacts clinical trust and adoption. Vision Transformers (ViTs) have demonstrated state-of-the-art performance in diagnostic imaging; however, their complex attention mechanisms pose challenges to explainability. This study evaluates the explainability of different Vision Transformer architectures and pre-training strategies - ViT, DeiT, DINO, and Swin Transformer - using Gradient Attention Rollout and Grad-CAM. We conduct both quantitative and qualitative analyses on two medical imaging tasks: peripheral blood cell classification and breast ultrasound image classification. Our findings indicate that DINO combined with Grad-CAM offers the most faithful and localized explanations across datasets. Grad-CAM consistently produces class-discriminative and spatially precise heatmaps, while Gradient Attention Rollout yields more scattered activations. Even in misclassification cases, DINO with Grad-CAM highlights clinically relevant morphological features that appear to have misled the model. By improving model transparency, this research supports the reliable and explainable integration of ViTs into critical medical diagnostic workflows.

---

### Asking Clarifying Questions for Preference Elicitation With Large Language Models

Authors: Ali Montazeralghaem, Guy Tennenholtz, Craig Boutilier, Ofer Meshi
Published: 2025-10-13
Categories: cs.AI
arXiv: https://arxiv.org/abs/2510.12015v1

Large Language Models (LLMs) have made it possible for recommendation systems to interact with users in open-ended conversational interfaces. In order to personalize LLM responses, it is crucial to elicit user preferences, especially when there is limited user history. One way to get more information is to present clarifying questions to the user. However, generating effective sequential clarifying questions across various domains remains a challenge. To address this, we introduce a novel approach for training LLMs to ask sequential questions that reveal user preferences. Our method follows a two-stage process inspired by diffusion models. Starting from a user profile, the forward process generates clarifying questions to obtain answers and then removes those answers step by step, serving as a way to add ``noise'' to the user profile. The reverse process involves training a model to ``denoise'' the user profile by learning to ask effective clarifying questions. Our results show that our method significantly improves the LLM's proficiency in asking funnel questions and eliciting user preferences effectively.

---

### Embedding the Teacher: Distilling vLLM Preferences for Scalable Image Retrieval

Authors: Eric He, Akash Gupta, Adian Liusie, Vatsal Raina, Piotr Molenda, Shirom Chabra, Vyas Raina
Published: 2025-10-13
Categories: cs.IR, cs.LG
arXiv: https://arxiv.org/abs/2510.12014v1

Text--image retrieval is necessary for applications such as product recommendation. Embedding-based approaches like CLIP enable efficient large-scale retrieval via vector similarity search, but they are primarily trained on literal caption-like text--image pairs and often fail to capture abstract or persona-driven attributes common in product recommendation applications (e.g., ``a gift for a mother who loves gardening''). In contrast, state-of-the-art vision--language models (vLLMs) can align text with images in a flexible manner, but their limited context window prevents them from directly handling retrieval over large catalogs. We propose a framework that distills the preference rankings of a powerful vLLM into an embedding-based system, transferring its nuanced alignment abilities while maintaining the inference-time scalability of an embedding-based approach. Experiments on persona-driven product recommendation tasks demonstrate that our method significantly outperforms existing embedding-based baselines, providing an efficient solution for personalized text--image retrieval.

---

### Statistical Guarantees for High-Dimensional Stochastic Gradient Descent

Authors: Jiaqi Li, Zhipeng Lou, Johannes Schmidt-Hieber, Wei Biao Wu
Published: 2025-10-13
Categories: stat.ML, cs.LG
arXiv: https://arxiv.org/abs/2510.12013v1

Stochastic Gradient Descent (SGD) and its Ruppert-Polyak averaged variant (ASGD) lie at the heart of modern large-scale learning, yet their theoretical properties in high-dimensional settings are rarely understood. In this paper, we provide rigorous statistical guarantees for constant learning-rate SGD and ASGD in high-dimensional regimes. Our key innovation is to transfer powerful tools from high-dimensional time series to online learning. Specifically, by viewing SGD as a nonlinear autoregressive process and adapting existing coupling techniques, we prove the geometric-moment contraction of high-dimensional SGD for constant learning rates, thereby establishing asymptotic stationarity of the iterates. Building on this, we derive the $q$-th moment convergence of SGD and ASGD for any $q\ge2$ in general $\ell^s$-norms, and, in particular, the $\ell^{\infty}$-norm that is frequently adopted in high-dimensional sparse or structured models. Furthermore, we provide sharp high-probability concentration analysis which entails the probabilistic bound of high-dimensional ASGD. Beyond closing a critical gap in SGD theory, our proposed framework offers a novel toolkit for analyzing a broad class of high-dimensional learning algorithms.

---

### Generate Logical Equivalence Questions

Authors: Xinyu Wang, Haoming Yu, Yicheng Yang, Zhiyuan Li
Published: 2025-10-13
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.12001v1

Academic dishonesty is met with zero tolerance in higher education, yet plagiarism has become increasingly prevalent in the era of online teaching and learning. Automatic Question Generation (AQG) presents a potential solution to mitigate copying by creating unique questions for each student. Additionally, AQG can provide a vast array of practice questions. Our AQG focuses on generating logical equivalence questions for Discrete Mathematics, a foundational course for first-year computer science students. A literature review reveals that existing AQGs for this type of question generate all propositions that meet user-defined constraints, resulting in inefficiencies and a lack of uniform question difficulty. To address this, we propose a new approach that defines logical equivalence questions using a formal language, translates this language into two sets of generation rules, and develops a linear-time algorithm for question generation. We evaluated our AQG through two experiments. The first involved a group of students completing questions generated by our system. Statistical analysis shows that the accuracy of these questions is comparable to that of textbook questions. The second experiment assessed the number of steps required to solve our generated questions, textbook questions, and those generated by multiple large language models. The results indicated that the difficulty of our questions was similar to that of textbook questions, confirming the quality of our AQG.

---

### UALM: Unified Audio Language Model for Understanding, Generation and Reasoning

Authors: Jinchuan Tian, Sang-gil Lee, Zhifeng Kong, Sreyan Ghosh, Arushi Goel, Chao-Han Huck Yang, Wenliang Dai, Zihan Liu, Hanrong Ye, Shinji Watanabe, Mohammad Shoeybi, Bryan Catanzaro et al.
Published: 2025-10-13
Categories: cs.SD, cs.CL, cs.LG
arXiv: https://arxiv.org/abs/2510.12000v1

Recent advances in the audio language modeling (ALM) domain tackle audio understanding and text-to-audio generation as separate tasks. Very few studies attempt to unify these tasks -- an essential step toward advanced multimodal reasoning. This paper introduces U}nified Audio Language Model (UALM), which aims to unify audio understanding, text-to-audio generation, and multimodal reasoning in a single model. To achieve this goal, we first present UALM-Gen, a text-to-audio language model that directly predicts audio tokens and is comparable to state-of-the-art diffusion-based models. We then demonstrate, using proper data blending, training recipes, and inference techniques, that our single UALM model matches the quality of state-of-the-art specialized models in audio understanding, text-to-audio generation, and text reasoning. Furthermore, we present UALM-Reason, a multimodal reasoning model that utilizes both text and audio in the intermediate thinking steps to facilitate complex generation tasks. To our knowledge, this is the first demonstration in audio research of cross-modal generative reasoning, with its effectiveness confirmed by subjective evaluations.

---

### SAGE: A Top-Down Bottom-Up Knowledge-Grounded User Simulator for Multi-turn AGent Evaluation

Authors: Ryan Shea, Yunan Lu, Liang Qiu, Zhou Yu
Published: 2025-10-13
Categories: cs.CL
arXiv: https://arxiv.org/abs/2510.11997v3

Evaluating multi-turn interactive agents is challenging due to the need for human assessment. Evaluation with simulated users has been introduced as an alternative, however existing approaches typically model generic users and overlook the domain-specific principles required to capture realistic behavior. We propose SAGE, a novel user Simulation framework for multi-turn AGent Evaluation that integrates knowledge from business contexts. SAGE incorporates top-down knowledge rooted in business logic, such as ideal customer profiles, grounding user behavior in realistic customer personas. We further integrate bottom-up knowledge taken from business agent infrastructure (e.g., product catalogs, FAQs, and knowledge bases), allowing the simulator to generate interactions that reflect users' information needs and expectations in a company's target market. Through empirical evaluation, we find that this approach produces interactions that are more realistic and diverse, while also identifying up to 33% more agent errors, highlighting its effectiveness as an evaluation tool to support bug-finding and iterative agent improvement.

---

### Prompt-Guided Spatial Understanding with RGB-D Transformers for Fine-Grained Object Relation Reasoning

Authors: Tanner Muturi, Blessing Agyei Kyem, Joshua Kofi Asamoah, Neema Jakisa Owor, Richard Dyzinela, Andrews Danyo, Yaw Adu-Gyamfi, Armstrong Aboah
Published: 2025-10-13
Categories: cs.CV
arXiv: https://arxiv.org/abs/2510.11996v1

Spatial reasoning in large-scale 3D environments such as warehouses remains a significant challenge for vision-language systems due to scene clutter, occlusions, and the need for precise spatial understanding. Existing models often struggle with generalization in such settings, as they rely heavily on local appearance and lack explicit spatial grounding. In this work, we introduce a dedicated spatial reasoning framework for the Physical AI Spatial Intelligence Warehouse dataset introduced in the Track 3 2025 AI City Challenge. Our approach enhances spatial comprehension by embedding mask dimensions in the form of bounding box coordinates directly into the input prompts, enabling the model to reason over object geometry and layout. We fine-tune the framework across four question categories namely: Distance Estimation, Object Counting, Multi-choice Grounding, and Spatial Relation Inference using task-specific supervision. To further improve consistency with the evaluation system, normalized answers are appended to the GPT response within the training set. Our comprehensive pipeline achieves a final score of 73.0606, placing 4th overall on the public leaderboard. These results demonstrate the effectiveness of structured prompt enrichment and targeted optimization in advancing spatial reasoning for real-world industrial environments.

---