The paper introduces an uncertainty-guided progressive contraction framework to reduce the amount of reasoning supervision needed for large language models. By building Monte Carlo reasoning trees around intermediate states, the method contracts the training corpus into a compact teacher-informed pool and then performs learner-aware contraction during supervised fine-tuning. Experiments show that the approach can use as little as 0.22% of the optimization budget of full-data SFT while maintaining or improving reasoning performance across diverse benchmarks and model families.
By Suorong Yang, Jiaxin Wang, Jiaming Su, Chentao Yue, Furao Shen, Peng Ye, Soujanya Poria
The paper introduces TQS, a Trajectory Quality Score that evaluates coding-agent interaction trajectories based on structural signals such as information focus, intent–action alignment, and policy dynamics, rather than just final outcomes. TQS operates in linear time without requiring an LLM judge and is applied to 155,144 successful trajectories from a public corpus. Fine‑tuning agents on trajectories ranked highest by TQS consistently outperforms models trained on random or low‑ranked subsets on the SWE‑Bench‑Verified benchmark, demonstrating the value of process‑level quality filtering.
By Sida Li, Yuanxin Liu, Shicheng Li, Linli Yao, Xu Sun, Haoli Bai, Xiaohui Li
The paper introduces Artisan, a multi‑agent pipeline that builds on Articraft to generate physically grounded articulated 3D assets. Artisan incorporates physical simulation feedback to ensure attributes such as collision‑free motion, stability under gravity, and accurate mass and material properties. The authors also present a unified benchmark covering spatial fidelity, physical plausibility, downstream task performance, and overall quality, evaluated on 100 reference assets and 100 stability prompts, where Artisan shows competitive spatial fidelity and the lowest collision and highest stability rates among tested models.
By Xudong Cai, Ruijie Zhu, Haoliang Shang, Ruining Li, Marc Pollefeys, Shangzhe Wu, Andrea Vedaldi, Chuanxia Zheng
The paper introduces ChemDiffAgent, a masked diffusion language agent designed for tool‑integrated chemical reasoning. Unlike traditional autoregressive agents, it treats reasoning as iterative denoising over interleaved reasoning and tool‑use trajectories, allowing joint decision making under full‑trajectory context. A two‑stage post‑training framework further refines tool use and reasoning, and the authors benchmark the model on six chemistry tasks, reporting a ~20% performance gain over autoregressive baselines and competitive results against larger models.
By Mengdi Liu, Chenghao Jia, Hong Chang, Shiguang Shan
COFFEQA introduces a counterfactual-guided few‑shot distillation method for open‑book question answering, leveraging minimally modified contexts that prompt the teacher model to produce alternative answers. By pairing factual samples with validated counterfactuals—filtered using a new Context Following Difficulty metric—COFFEQA constrains the student’s decision boundary to better match the teacher’s. Experiments on multiple QA benchmarks show that COFFEQA consistently outperforms factual‑only distillation, achieving significant gains in EM and F1 scores with a limited few‑shot budget.
By Kavinda Kehelella, Sanghamitra Dutta
The paper introduces the Answer-Evidence Support Protocol (AESP) to evaluate whether the grounding evidence in remote sensing visual question answering actually supports the model’s answer. It applies task‑specific rules for existence, counting, and comparison questions and is tested on the newly constructed G‑RSVQA benchmark, which contains 15,027 samples with detailed object‑level annotations. Experiments on seven grounded vision‑language models show that up to 50.82% of correct answers are not supported by the provided grounding evidence.
By Ping Guo, Zhiqi Huang, Xinran Li, Linkun Yu, Jingyin Pei, Kamata Seiichiro
Neurons++ is a visual‑stream framework that reconstructs both static images and dynamic videos from fMRI data. It builds on the earlier Neurons model by extending from video‑only to image‑and‑video reconstruction and replacing parallel task conditioning with a serial, cortex‑inspired decoding pipeline. The approach first pre‑trains on image‑fMRI pairs for stable spatial‑semantic initialization, then fine‑tunes separately for image and video tasks, organizing fMRI representations into spatial, semantic, language‑level, and temporal cues, and achieves state‑of‑the‑art video reconstruction with competitive image results while enabling neural‑to‑visual interpretability and ROI‑guided semantic modulation.
By Haonan Wang, Yuchen Wang, Jingyu Lu, Qixiang Zhang, Lehan Wang, Xuanqi Huang, Xiaomeng Li
The study evaluates multiple computational methods for detecting dead cells in fluorescence microscopy images, comparing heuristic thresholding, classification with numerical and image features, and anomaly detection techniques. Using a dataset of 258 images and 65,527 cells, the best performance was achieved by deep learning image classifiers (ResNet-18) and traditional numerical classifiers (XGBoost), outperforming manual heuristics and anomaly detection. The results highlight the difficulty of reliably identifying apoptotic cells despite advanced machine learning approaches.
By Prateek Verma, Benjamin Butter, Jian Zhang, Xintao Wu
The paper introduces a new ellipsoid‑specific framework for camera pose estimation from a single ellipse‑ellipsoid correspondence. It shows that the problem can be reduced to either a position‑only or orientation‑only estimation, with the remaining unknowns solvable in closed form, and further to a single degree‑of‑freedom problem with an analytical pose solution. The authors illustrate the theory with visual examples, discuss practical aspects, and provide source code to aid efficient pose estimation.
By Vincent Gaudilli\`ere, Gilles Simon, Marie-Odile Berger
arXiv:2610. 13461v1 Announce Type: new Abstract: AI systems are making rapid progress in mathematics, but this progress is hard to measure.
By Pablo Manrique, Hai Xia, Stefan Szeider
MAGMA-Bench is an architecture‑agnostic benchmark designed to evaluate long‑horizon embodied agents in dynamic multi‑robot environments. It tests agents across increasing task horizons and introduces controlled online perturbations such as execution failures, task interruptions, and mission updates, while requiring coordination among robots with heterogeneous capabilities. The benchmark separates task‑level reasoning from perception and low‑level control through a structured skill‑level interface and provides 261,646 task‑level training examples along with an expert‑trace generation pipeline.
By Loan Bernat (LAAS-GEPETTO), Matthieu Grard (LAAS-RAP), Ariane Herbulot (LAAS-RAP), Florent Lamiraux (LAAS-GEPETTO)
The paper "Can Language Models Play the Wikipedia Game?" investigates how language models perform on an information‑retrieval task where players navigate from one Wikipedia article to a target using only article links. The authors compare sentence‑embedding models and large language models across five Wikipedia languages and five academic‑literature datasets, building extensive local graphs of Wikipedia pages and paper citations with OpenAlex. They benchmark these models against human and TF‑IDF baselines, finding that no model significantly outperforms humans, though the best models (GPT‑o3 mini, GPT‑4o, QwQ:32B, and DeepSeek‑R1:70B) achieve comparable success, with performance varying by language and topic.
By Alexander Hickey, Joanna Knight, James Bishop
arXiv:2604. 09434v3 Announce Type: replace-cross Abstract: Turbulent boundary layers over aerodynamic surfaces are a major source of aircraft drag, yet their control remains challenging due to multiscale dynamics and spatial variability, particularly under adverse pressure gradients.
By Yuning Wang, Pol Suarez, Mathis Bode, Ricardo Vinuesa
The paper introduces w$^{2}$VLA, a Vision‑Language‑Action model that restructures information flow to separate declarative knowledge (concepts and entity semantics) from procedural knowledge (how to act). By modulating the robot state sequence with visual, spatial, and skill information in a compositional, interpretable way, the model achieves robust behavior cloning and unprecedented zero‑shot skill transfer to unseen objects, overcoming brittleness seen in current state‑of‑the‑art VLA models.
By Nikolaos Tsagkas, Andreas Sochopoulos, Chris Xiaoxuan Lu, Oisin Mac Aodha, Alexandros Kouris
ThinkNet is a validation‑controlled framework that uses train‑only normalization, evolutionary search guided by validation, and validation‑gated inference to build compact motor‑imagery EEG decoders that generalize to unseen subjects. In a fixed benchmark on the BCI Competition IV‑2a dataset, the best single model achieved 44.35 % accuracy with only 4.9 k parameters, while a compact ensemble of six families reached 43.31 % accuracy. Across a broader search, models with ≤25 k parameters outperformed larger ones after retraining, but validation reliability remains a bottleneck under subject shift.
By Abdul Basit, Saim Rehman, Muhammad Shafique
AgroGround is a large‑scale dataset that links plant disease and agricultural target labels to image regions, enabling models to both recognize and localize these targets. It was created by automatically converting eight agricultural VQA datasets into 794,850 grounded instruction examples, including healthy images for negative supervision. Fine‑tuning a shared vision‑language model on these instructions improves recognition, joint accuracy, and healthy‑image abstention, outperforming its annotation teacher on both the AgroGround benchmark and the external PlantSeg test set.
By Abdulla Alshehhi, Zongyan Han, Rao Muhammad Anwer
Multi-Stream Attention Residuals (MSAR) extends the Attention Residuals (AttnRes) mechanism by introducing multi-head depth attention, inspired by multi-head attention along the sequence axis. MSAR generates diverse streams from sequence context via multi-scale causal convolutions, then adaptively mixes them to produce a single-stream layer input, enabling multi-stream computation without extra storage or communication. Across 12 downstream benchmarks at 18B and 28B MoE scales, MSAR consistently outperforms prior residual variants and achieves a 1.58× iso-loss compute speedup over the standard residual baseline.
JitterVLA is a replay benchmark designed to evaluate motion quality in vision‑language‑action (VLA) driven human‑robot manipulation. It replays identical command sequences and arrival traces through different controllers to isolate controller effects, testing on synthetic reaches, hand trajectories from 20 people, three open VLA pipelines (838 episodes), and a physical UR3 robot. The benchmark shows that interpolation reduces modeled p99 jerk by 83% for human motion with a 200 ms delay, and that jerk‑limited generation further lowers jerk at a higher cost, while entropy‑based methods yield mixed results across pipelines.
Agent4Spec is an autonomous multi‑agent system that uses large language models to continuously improve API specifications. It includes a validation pipeline that detects and corrects OpenAPI syntax errors, and a market‑analysis pipeline that researches competitors, finds functional gaps, and auto‑generates missing API features. In tests on two curated datasets, the best LLM achieved 100 % syntax error detection, over 95 % success in implementing corrections, and more than 50 % adoption of suggested endpoint improvements.
The paper introduces Data-Adaptive Mahalanobis Cross-Head Attention (DAMCHA), a new attention mechanism that extends Multi-Head Attention by learning a full cross-head metric matrix, thereby capturing inter-head interactions beyond the block-diagonal structure of standard MHA. DAMCHA parameterizes this metric as an input-dependent function, allowing the attention geometry to adapt dynamically to the data’s intrinsic manifold structure. The authors also propose regularization techniques, such as stack-wise parameter sharing, to maintain computational efficiency and stable optimization, and demonstrate that DAMCHA-equipped Transformers outperform existing baselines on multiple benchmark tasks.