Introduces 'sharpness dimension' to explain improved generalization at the edge of stability.

Mario Tuci, Caner Korkmaz, Umut Şimşekli et al.

Study fluctuations of functionals in infinite-width random neural networks on spheres, revealing three distinct limiting behaviors.

Simmaco Di Lillo, Leonardo Maini, Domenico Marinucci

Proposes Safe EWC and CF-EWC algorithms for safe continual reinforcement learning in non-stationary environments.

Austin Coursey, Abel Diaz-Gonzalez, Marcos Quinones-Grueiro et al.

FASTER method reduces computational cost by early action sample filtering during denoising while maintaining RL performance.

Perry Dong, Alexander Swerdlow, Dorsa Sadigh et al.

VLA Foundry: A unified framework for training Vision-Language-Action models, enhancing multi-task tabletop manipulation policies.

Jean Mercat, Sedrick Keh, Kushal Arora et al.

Adversarial training enables Vision Transformers to achieve near-zero robust training loss and robust generalization error under moderate perturbation budgets.

Jiaming Zhang, Meng Ding, Shaopeng Fu et al.

Discovering a shared logical subspace in LLMs improves logical reasoning accuracy by up to 11% via alignment of natural-language and symbolic views.

Feihao Fang, My T. Thai, Yuanyuan Lei

Evaluates a VPP dispatch algorithm in smart distribution systems using a co-simulation framework, revealing significant impacts of communication delays.

Houchao Gan

Utilizing determinantal point processes for Monte Carlo integration to enhance estimator variance convergence speed.

Guillaume Gautier, Rémi Bardenet, Michal Valko

A-MAR framework enhances multimodal art retrieval explanation quality through structured reasoning plans.

Shuai Wang, Hongyi Zhu, Jia-Hong Huang et al.

InsightGen generates diverse and relevant insights to enhance open-ended document QA.

Saransh Sharma, Pritika Ramu, Aparna Garimella et al.

Mask World Model predicts semantic masks instead of pixels, enhancing robust robot policy learning, excelling in LIBERO and RLBench.

Yunfan Lou, Xiaowei Chi, Xiaojie Zhang et al.

MATCH method improves peg-in-hole task success rate by 35% under high noise, reducing average force by 30%.

Hunter L. Brown, Geoffrey Hollinger, Stefan Lee

RAPIDDS framework enhances human-robot teaming efficiency through multi-cycle spatio-temporal adaptation, significantly improving plan fluency and user preference.

Alex Cuellar, Michael Hagenow, Julie Shah

ECLASS-augmented dense retrieval method achieves 94.3% HitRate@5 in semantic search for electronic components.

Nico Baumgart, Markus Lange-Hegermann, Jan Henze

GPT models predict experience ratings from open-ended survey text; prompt optimization improves accuracy by 2%.

Andrew Hong, Jason Potteiger, Luis E. Zapata

Gesture recognition using OpenCLIP visual learning model improves AcoustoBot swarm interaction accuracy to 87.8%.

Alex Lin, Lei Gao, Narsimlu Kemsaram et al.

Micro Language Models (μLMs) enable instant responses by generating the first 4-8 words on-device, with cloud models completing the response.

Wen Cheng, Tuochao Chen, Karim Helwani et al.

SafetyALFRED evaluates safety planning in multimodal LLMs in kitchen settings, finding good hazard recognition but low risk mitigation success.

Josue Torres-Fonseca, Naihao Deng, Yinpei Dai et al.

The ESKF-PRE-VMPC framework reduces RMSE by 52.63% and 75.04% in UAV pipeline inspection without wind.

Wen Li, Hui Wang, Jinya Su et al.