Unified framework for LLM post-training based on trajectory provenance, supporting expansion, reshaping, and consolidation roles.

Shiwan Zhao, Zhihu Wang, Xuyang Zhao et al.

Proposes semantic UI element injection attack using black-box methods, overlaying harmless icons to mislead GUI agents, boosting success rate 3.5-6.9×

Wenkui Yang, Chao Jin, Haisu Zhu et al.

RoboAgent employs capability chaining within a single VLM, decomposing complex embodied tasks into basic vision-language problems, enhancing long-term planning.

Peiran Xu, Jiaqi Zheng, Yadong Mu

Proposes a training-free open-vocabulary semantic segmentation method using analytic solutions based on distribution discrepancy analysis, achieving SOTA on 8 datasets.

Jiahao Li, Yang Lu, Yachao Zhang et al.

ConfigSpec framework optimizes distributed inference configuration by analyzing edge devices and draft models.

Xiangchen Li, Saeid Ghafouri, Jiakun Fan et al.

EgoVerse unifies large-scale egocentric human demonstration data, enabling cross-embodiment robot learning with 80k episodes, 1965 tasks, and multi-institution collaboration.

Ryan Punamiya, Simar Kareer, Zeyi Liu et al.

TAMEn enhances bimanual manipulation success from 34% to 75% through tactile-aware closed-loop data collection.

Longyan Wu, Jieji Ren, Chenghang Jiang et al.

Introduced PAMELA dataset and personalized reward model, improving individual preference prediction in T2I generation.

Anne-Sofie Maerten, Juliane Verwiebe, Shyamgopal Karthik et al.

PhyEdit combines explicit geometric simulation with Diffusion Transformer to achieve high-precision 3D-aware image manipulation, outperforming existing methods with significant improvements in geometric accuracy and physical consistency.

Ruihang Xu, Dewei Zhou, Xiaolong Shen et al.

This study evaluates multilingual embedding models' ability to encode general proficiency, revealing poor cross-corpus transfer with performance dropping from QWK≈0.7 to 0.38.

Laurits Lyngbaek, Ross Deans Kristensen-McLachlan

ModuSeg decouples object discovery and semantic retrieval, enabling training-free weakly supervised segmentation with state-of-the-art performance.

Qingze He, Fagui Liu, Dengke Zhang et al.

MENO combines improved MeanFlow with neural operators for high-resolution dynamical system prediction, achieving 2× spectral accuracy and 14× faster inference.

Tianyue Yang, Xiao Xue

Introduced CLI-Tool-Bench, a black-box differential testing framework, achieving only 43.8% success in end-to-end CLI tool generation, highlighting ongoing challenges.

Ruida Hu, Xinchen Wang, Chao Peng et al.

Proposes DOC-GS framework combining depth-guided Dropout and Dark Channel Prior for reliable sparse-view 3D Gaussian reconstruction.

Hantang Li, Qiang Zhu, Xiandong Meng et al.

Proposed an expert-wise mixed precision quantization method based on router l2 norm changes, improving model accuracy and inference efficiency.

Mohammed Nowaz Rabbani Chowdhury, Kaoutar El Maghraoui, Hsinyu Tsai et al.

Proposes CMRM, a quantile-calibrated framework that enhances robustness under label noise without prior knowledge, improving accuracy by up to 3.39%.

Yuanjie Shi, Peihong Li, Zijian Zhang et al.

This paper systematically reviews video generation evolution from GANs to diffusion and autoregressive models, analyzing core principles and innovations.

Teng Hu, Jiangning Zhang, Hongrui Huang et al.

AgentCE-Bench evaluates agents with scalable horizons and controllable difficulty using hidden slots H and decoy budget B.

Wang Yang, Chaoda Song, Xinpeng Li et al.

BOSCH employs black-box binary optimization via large neighborhood search for short-context head selection, outperforming static heuristics across models from 1.7B to 30B parameters.

Abbas Ghaddar, Ivan Kobyzev, Boxing Chen et al.

Constructed a one-dimensional Noetherian local domain D, using Elliott's flatness criterion, to show Int(D) is not flat as a D-module, providing a counterexample to general flatness assumptions.

Haotian Ma