LongZero is a reinforcement learning method that enables large language models to improve long‑context reasoning without external data. It works by having the model first generate a compact, structured question‑answer pair with supporting evidence, then expanding that evidence into a long, multi‑document context. The approach combines self‑written multi‑hop reasoning and multi‑round retrieval tasks, trained with GRPO, and outperforms baseline RL methods on five long‑context reasoning benchmarks.
By Zhennan Wan, Jianfei Chen
TraceForge introduces a two‑stage training method that generates reasoning traces in the style of a target model family, using supervised learning followed by adversarial refinement. The system produces traces that are judged as authentic with scores comparable to real traces, and when applied to teacher answers, it improves performance on coding benchmarks such as HumanEval+ and MBPP+. The approach addresses the limitation of existing models that only return answers without detailed reasoning, enabling more transparent and learnable outputs.
By Yan Wang
The paper introduces Weak-form Anchored Rollout for PDEs (WARP), a method that learns PDE solvers by differentiating through multistep rollouts anchored at observed states. WARP addresses noise sensitivity and data consistency issues of sparse regression approaches, producing numerically stable rollouts and improved coefficient recovery across eight fluid and plasma benchmarks, including high‑dimensional turbulent simulations. The method is tailored to systems solvable by spectral Galerkin techniques and outperforms a single‑step weak‑form SINDy baseline in rollout accuracy and statistical metrics.
By Mellow Baixuan Chen, Alan A. Kaptanoglu
The paper introduces Sketch2Sim, an end‑to‑end modular framework that transforms hand‑drawn circuit diagrams into SPICE‑compatible netlists and automatically computes nodal voltages and currents. It employs a three‑stage pipeline: dual YOLO detectors for components and text, a CRNN‑CTC OCR module for symbol value extraction, and image‑processing techniques (thresholding, skeletonization, DBSCAN) to identify nodes and map components. Experiments on a curated dataset of 1,000 RLC circuits show robust performance across diverse handwriting styles, lighting, and clutter, and the system includes a user interface for rapid digitization and simulation.
By Mohammad Kawsar, Al Amin Islam Ridoy, Md Rakib Hossain, Farzana Akter, AKM Bahalul Haque, Rejwan Bin Sulaiman, Mohammed Amin Almaiah
SkillEval introduces a new evaluation framework for large language models that moves beyond single scalar scores to perform cross-model, skill-level analysis. It automatically discovers interpretable skills from benchmark questions using LLM meta‑cognition and applies cognitive diagnosis models to estimate per‑skill abilities. Experiments on 3,811 LLMs across five benchmarks show that SkillEval yields more calibrated, informative model profiles, enabling tasks such as efficient LLM routing and adaptive testing.
By Berke Arda, Peng Cui, Qiaoyuan Zheng, Rudolf Debelak, Mubashara Akhtar, Mrinmaya Sachan
The study investigates when adding zero‑shot pseudo‑labels from vision‑language models (VLMs) to small expert‑labelled medical imaging datasets ceases to improve, and may even degrade, classifier performance. Across three benchmarks—PatchCamelyon, ISIC 2019, and NIH ChestX‑ray14—the crossover point occurs after roughly 20–50, 100, and 250–500 expert labels, respectively, with additional pseudo‑labels reducing macro‑AUC by up to 0.14 points. The analysis shows that the VLM’s labeling quality, rather than the student model size, limits performance, and a simple bias‑variance rule can predict the benefit of pseudo‑labels in most settings.
By Bruce Changlong Xu, Jose K. James, Alexander J. Ryu
The paper introduces the Register Anything Model (RAM), a unified approach for 3D point cloud registration that can robustly estimate 6‑DoF transformations across a wide range of domains, including indoor, outdoor, and object scenes. RAM overcomes domain‑specific challenges by employing a geometry‑aware normalization that voxelizes and standardizes point clouds based on their geometric properties, followed by a hierarchical coarse‑to‑fine pipeline with a geometric transformer that focuses on local consistency. Experiments on 12 benchmarks show that RAM achieves state‑of‑the‑art inlier ratios and registration recall, outperforming methods trained on individual domains and demonstrating strong generalization.
By Zheng Qin, Junhua Xi, Yan Huang, Jiahong Lai, Anwen Huang, Qiong Li, Guangda Zhang, Kai Xu
EPIC introduces a new PPO-style clipping method that operates at the prefix level rather than token level, using the exact reverse-KL coordinate to regulate policy drift during autoregressive generation. By reformulating generation as a finite-horizon Markov decision process, the authors derive a performance-difference identity that identifies the cumulative likelihood ratio of the generated prefix as the exact importance weight for each policy-improvement step. Experiments on off‑policy reinforcement learning with a sparse mixture‑of‑experts language model demonstrate that EPIC consistently outperforms standard and sequence‑level clipping baselines on mathematical‑reasoning benchmarks while remaining competitive with routing‑replay alternatives.
By Chengxi Li, Hanlin Wu, Zhicheng Cai, Huan-ang Gao, Wei-Ying Ma, Jingjing Liu, Hao Zhou
The paper presents a multi‑label sentiment analysis framework that follows Plutchik’s wheel of emotions and is applicable to 100 languages. It introduces a dynamic weighting scheme to balance class contributions during training, adapts focal loss for harder instances, and derives a linear‑time method for selecting optimal class‑specific thresholds to maximize macro‑F1. Experiments show state‑of‑the‑art performance across multiple languages and metrics, and the authors release the code publicly.
By Selim F. Yilmaz, E. Batuhan Kaynak, Aykut Ko\c{c}, Hamdi Dibeklio\u{g}lu, Suleyman S. Kozat
The study examines how the choice of camera—Canon versus Nikon—affects image quality and AI performance in retinal screening for diabetic retinopathy. In a Brazilian cohort of 8,524 patients, Nikon images were more than twice as likely to be of inadequate quality compared with Canon, despite being cheaper. Using a frozen pretrained retinal foundation model, the authors evaluated 12 disease classification tasks across different resolutions and classifiers, finding that camera-related performance differences vary widely and are not consistently mitigated by higher resolution or classifier choice.
By Lisa Y. W. Tang
arXiv:2610.29107v1 Announce Type: new
Abstract: Recovering dynamic 3D scene geometry, appearance, and the physical motion states that govern scene evolution from multi-view videos is important yet ch...
By Shangjia Liu, Jinxi Li, Siyuan Zhou, Bo Yang, Bing Wang
arXiv:2610.29176v1 Announce Type: new
Abstract: In video action detection, modeling the temporal dependencies in continuous video streams is crucial. Previous methods based on deep neural networks ca...
By Yicheng Qiu, Keiji Yanai
ProactEgoMem is a benchmark for evaluating proactive memory in streaming egocentric video, focusing on three tasks—Information Capture, State Tracking, and Location Tracking—to reflect real-world environmental needs. It introduces two levels of proactiveness: Notification Level, which requires a single timely alert, and Intervention Level, which demands continuous memory maintenance via an active tool lifecycle. The benchmark includes daily and scientific datasets with corresponding QA sets, and shows that current models struggle to capture memory both timely and accurately.
By Ruotong Liao, Meikel Kokowski, Xiaoxiao Sun, Yongqin Xian, Laura Bravo S\'anchez, Shiye Su, Federico Tombari, Serena Yeung-Levy, Volker Tresp
The report introduces pplx-embed-v2-late, a pair of multilingual and multimodal late‑interaction models sized at 0.6B and 9B parameters. Both models achieve state‑of‑the‑art performance on text, image, and visual document retrieval benchmarks, with the 0.6B model matching larger models on visual document tasks. They are trained via representation distillation from an 18B teacher, enabling a shared embedding space that allows documents to be indexed with the 9B model and queried with the low‑latency 0.6B model.
By Markus Krimmel, Louis Milliken, Ekaterina Glazkova, Maximilian Schall, Antoine Chaffin, Sedigheh Eslami, Jonas Golde, Denis Bykov, Claire Beamer, Jeremy Yang, Bo Wang
The paper introduces TF‑IG, a compact two‑branch model that learns heart‑rate from photoplethysmography by combining temporal pulse morphology and frequency‑domain periodicity. A novel interaction‑gated fusion module explicitly models agreement and discrepancy between the two domains, while a shared regression head maps the fused features to HR predictions. On the BIDMC PPG dataset, TF‑IG achieves a mean absolute error of 1.227 BPM and 92.28 % of estimates within 3 BPM using only 84.6 K parameters.
By Anh Khoi Pham, Phat Lam, Thien Truong Nguyen Ly, Lam Pham
ReAPO: Retry-Aware Policy Optimization for Efficient Exploration in LLM Reasoning proposes a new approach to improve language‑model reasoning by addressing the inefficiencies of independent stochastic rollouts. By framing credit assignment as a Stochastic Shortest Path problem, ReAPO modifies GRPO to allocate more generation to failed trajectories, reinforcing successful retries without rewarding failures. Experiments on Qwen3 and OLMo‑2 models across eight mathematical and scientific benchmarks show that ReAPO outperforms GRPO and other baselines, achieving higher first‑attempt accuracy and reducing generated tokens needed to reach peak performance.
By Nghi Dao, Calder Morton-Ferguson, Noah Golowich, Adam Block
The paper introduces RL-IF, a method for tracking non‑local training data influence in online reinforcement learning. It defines a trajectory‑level leave‑one‑out metric (RL‑LOO) and derives a first‑order estimator that accounts for both optimization and policy sampling effects, offering theoretical accuracy guarantees. Using RL‑IF, the authors develop RINSE, a data filtering technique that improves training efficiency and final performance on standard control tasks and an RLHF‑style toxicity‑mitigation scenario.
By Shixuan Liu, Cheng Tang, Yuzheng Hu, Fan Wu, Han Zhao, Jiaqi W. Ma
The paper introduces TOP‑Gain, a new criterion for Random‑Forest wrapper feature selection that uses truncated out‑of‑bag probability gain to reduce noise in feature scoring. By truncating negative probability changes, TOP‑Gain cuts roughly two‑thirds of noise variance while keeping signal, leading to more reproducible feature rankings. In extensive benchmarks on 26 OpenML‑CC18 datasets, TOP‑Gain outperformed 13 of 14 baselines in ROC‑AUC, accuracy, and Macro‑F1, matched VSURF’s accuracy with far less computation, and produced more stable feature sets without sacrificing accuracy.
By Sejong Oh
The paper introduces GPSCS, a learning-based framework that uses genetic programming to evolve a symbolic scoring function for selecting cutting planes in mixed-integer linear programming. It also employs a statistics-based method to decide how many top-ranked cuts to add during each separation round. Experiments on diverse synthetic and real-world benchmarks show that GPSCS reduces solving time by 6% over SCIP’s default strategy and outperforms a neural-network approach and a no-cut baseline, indicating strong generalization and robustness.
By Gwen maudet, Mathis Da Cunha, Gr\'egoire Danoy
The paper introduces a Bayesian Evolutionary Optimization framework for automatically discovering task‑specific multi‑agent system (MAS) topologies. It combines evolutionary operators that generate diverse workflow graphs with a graph‑aware Gaussian Process and Expected Improvement to guide costly evaluations toward promising candidates. Experiments on BigCodeBench for code generation and on tabular preprocessing benchmarks show that the evolved MAS architectures outperform other evaluated agent systems and improve performance on temporal datasets while remaining competitive on IID data.
By Dominik Jehle, Tom Zehle