SafeHarness enhances robot manipulation safety with obstacle-aware planning, achieving 71.9% task success and 87.5% collision avoidance.

Bingxin Xu, Yuzhang Shang, Zhen Dong et al.

Study finds embedding models weakly reflect physical measurements, influenced by superficial string similarity.

Juri Opitz, Andrianos Michail

Introduced Workspace Models for lightweight robotic memory via saliency-driven supervision, achieving a 91.5% task success rate.

Nitish Dashora, Douglas Chen, Idan Shenfeld et al.

PersistBench evaluates 4D models' visual memory using 360° videos, revealing short-term consistency issues.

Guangzhao He, Hadar Averbuch-Elor, Wei-Chiu Ma

SplashSplat reconstructs splashing liquids from multi-view videos, offering lower training costs and more plausible physical motion.

Peiyu Liu, Dingxi Zhang, Federico Tombari et al.

FAMOS predicts movable-part segmentation and joint parameters from sparse point clouds, improving segmentation performance by 64.5%.

Kevin Qu, Tao Sun, Massimiliano Viola et al.

Paint-Anything achieves any-color control for image generation/editing via shared hex prompts, improving ACBench-T2I by 85.3%.

Ji Xie, Dewei Zhou, Xinyu Huang et al.

ERCPMP-Gx dataset integrates endoscopic images, histopathology, and genomics for colorectal polyposis AI research.

Zahra Ghaffari, Massih Bahar, Mojgan Forootan et al.

Study explores how distribution shift affects pretraining gains in neural PDE surrogates using 254,909 RANS solutions.

Pochinapeddi Sai Bhargav, Nithin Somasekharan, Rohit Sunil Kanchi et al.

Using OverclaimBench, quantified LLM agents' overclaiming propensity; 67.9% didn't read all files.

Nolan Smyth, Yorguin-Jose Mantilla-Ramos, Pascal Jr Tikeng Notsawo et al.

Score Centering stabilizes off-policy RL by reducing training-inference mismatch drift.

Martin Marek, Max Ryabinin

Study improves coding agents' long-term performance via lightweight harness design; context management proves most effective.

Run-Ze Fan, Zihao Zhang, Simin Ma et al.

JEPA-Anything uses Orthogonal Predictive Factorization (OPF) to enhance cross-domain predictive models, improving prediction accuracy by 34.8%.

Taoyong Cui, Zhongyao Wang, Xinyue Xu et al.

PosteriorBench evaluates distributional accuracy of generative inverse solvers, revealing neural operators improve resolution robustness.

Jiachen Yao, Zi-Siang Hsu, Xi Deng et al.

StageGuard learns stage transitions for long-horizon robot tasks via agentic distillation, significantly improving prediction accuracy.

Jinbang Huang, Yuanzhao Hu, Zhiyuan Li et al.

RetireOPD method improves RL model success rate by 18.8% in ALFWorld using adaptive retirement strategy.

Yan Yu, Zhengxi Lu, Yizhou Liu et al.

Study shows gender bias in GPT models is transformed, not reduced, introducing 'harm laundering'.

Sarah Wyer, Sue Black, Noura Al Moubayed

GeoAAC improves real-world task success rate from 53.3% to 74.4% through geometry-based adaptive action chunking.

Xin Chen, Sen Chen, Yujuan Ding et al.

FlowSGS enhances inverse imaging accuracy using Split Gibbs Sampling and Stochastic Interpolants.

Tianao Li, Xinhui Qian, Emma Alexander

Using 1D CNN for RF fingerprinting under heterogeneous protocol interference, achieving up to 97% accuracy.

Tariq Abdul-Quddoos, Xiangfang Li, Lijun Qian