13 papers
GEAR: Granularity-Adaptive Advantage Reweighting for LLM Agents via Self-Distillation
Sijia Li, Yuchen Huang, Zifan Liu +7
Reinforcement learning has become a widely used post-training approach for LLM agents, where training commonly relies on outcome-level rewards that provide only coarse supervision.…
What to Ignore, What to React: Visually Robust RL Fine-Tuning of VLA Models
Yuanfang Peng, Jingjing Fu, Chuheng Zhang +6
Reinforcement learning (RL) fine-tuning has shown promise for Vision-Language-Action (VLA) models in robotic manipulation, but deployment-time visual shifts pose practical challeng…
Co-Evolving Latent Action World Models
Yucen Wang, Fengming Zhang, De-Chuan Zhan +3
Adapting pretrained video generation models into controllable world models via latent actions is a promising step towards creating generalist world models. The dominant paradigm ad…
Learning Additively Compositional Latent Actions for Embodied AI
Hangxing Wei, Xiaoyu Chen, Chuheng Zhang +5
Latent action learning infers pseudo-action labels from visual transitions, providing an approach to leverage internet-scale video for embodied AI. However, most methods learn late…
Discover, Learn, and Reinforce: Scaling Vision-Language-Action Pretraining with Diverse RL-Generated Trajectories
Rushuai Yang, Zhiyuan Feng, Tianxiang Zhang +6
Scaling vision-language-action (VLA) model pre-training requires large volumes of diverse, high-quality manipulation trajectories. Most current data is obtained via human teleopera…
What Do Latent Action Models Actually Learn?
Chuheng Zhang, Tim Pearce, Pushi Zhang +5
Latent action models (LAMs) aim to learn action-relevant changes from unlabeled videos by compressing changes between frames as latents. However, differences between video frames c…