SENSEI uses Vision Language Models to guide semantic exploration, enhancing RL behavior diversity.

Cansu Sancaktar, Christian Gumbsch, Andrii Zadaianchuk et al.

SePer measures retrieval utility via semantic perplexity reduction, showing high correlation with human judgments and outperforming traditional metrics.

Lu Dai, Yijie Xu, Jinhui Ye et al.

EPPO leverages evidential uncertainty to adapt quickly in non-stationary environments, outperforming state-of-the-art PPO variants in continuous control tasks.

Abdullah Akgül, Gulcin Baykal, Manuel Haußmann et al.

Object-aware video matting with cross-frame guidance achieves SOTA, reducing reliance on manual trimaps, with MAD 4.23 and MSE 0.31 on RVM.

Huayu Zhang, Dongyue Wu, Yuanjie Shao et al.

Phantom method trains robots using human videos, achieving 92% success without robot data.

Marion Lepert, Jiaying Fang, Jeannette Bohg

This study introduces the 'Safety Tax' phenomenon, showing safety alignment improves safety but degrades reasoning; validated via a simplified two-stage pipeline.

Tiansheng Huang, Sihao Hu, Fatih Ilhan et al.

Falcon reuses partial denoising actions based on historical data, achieving 2-7x speedup with negligible performance loss in diffusion policies.

Haojun Chen, Minghao Liu, Chengdong Ma et al.

GAS(P) evaluates gender bias in unprompted generated text, revealing persistent biases across models.

Jennifer Mickel, Maria De-Arteaga, Leqi Liu et al.

Statistical shape modeling reveals sex explains at least 25% of cardiac morphological variability.

Beatrice Moscoloni, Cameron Beeche, Julio A. Chirinos et al.

SEE-Net uses event data for adaptive brightness adjustment, enhancing image quality across broad light ranges.

Yunfan Lu, Xiaogang Xu, Hao Lu et al.

ACID employs transformer neural networks for conditional independence testing, achieving high accuracy and efficiency across diverse datasets.

Bao Duong, Nu Hoang, Thin Nguyen

DexGraspVLA achieves over 90% success in dexterous grasping in complex scenes.

Yifan Zhong, Xuchuan Huang, Ruochong Li et al.

Reveals three-phase transitions in LLMs using brain encoding, probing, and benchmark analyses.

Yuko Nakagi, Keigo Tada, Sota Yoshino et al.

EndoPBR uses physically-based differentiable rendering to estimate materials and lighting, enabling photorealistic novel view synthesis in surgical scenes.

John J. Han, Jie Ying Wu

Utilizing synthetic data and virtual simulation to identify and generate corner cases for enhanced autonomous vehicle safety.

Gabriel Kenji Godoy Shimanuki, Alexandre Moreira Nascimento, Lucio Flavio Vismari et al.

Q♯ algorithm uses distributional RL with KL regularization to optimize large language models post-training, achieving superior performance.

Jin Peng Zhou, Kaiwen Wang, Jonathan Chang et al.

Point Policy unifies observations and actions via key points, learning robot policies solely from offline human videos, achieving 75% performance boost.

Siddhant Haldar, Lerrel Pinto

InterMimic employs teacher-student RL framework to learn diverse, physically plausible human-object interactions from imperfect MoCap data, achieving zero-shot generalization.

Sirui Xu, Hung Yu Ling, Yu-Xiong Wang et al.

Symbolic mechanisms support abstract reasoning in LLMs; Llama-3.1 achieves 95% accuracy in rule induction tasks.

Yukang Yang, Declan Campbell, Kaixuan Huang et al.

This study uses translation chains to reveal iterative generation causes information distortion, with key metrics showing increased deviation over rounds.

Amr Mohamed, Mingmeng Geng, Michalis Vazirgiannis et al.