The paper introduces RoboImagine, a framework that refines autoregressive robotic world models by comparing candidate future trajectories using both semantic and visual feedback. It employs diverse search to generate varied candidate paths, a frozen multimodal evaluator for task‑relevant semantic preferences, and pixel‑level visual consistency checks to guide policy optimization. Experiments on RT‑1, BridgeV2, and RoboCasa show that RoboImagine improves visual fidelity (e.g., PSNR gains) and task relevance compared to standard fine‑tuning, with ablations confirming the benefits of diverse search and combined feedback signals.
By Hao Wu, Yuqi Li, Yizhou Zhao, Kun Wang, Xiao Luo, Yu Cheng, Bo An, Yingli Tian, Laszlo A. Jeni, Ming-Hsuan Yang
arXiv:2606. 15631v1 Announce Type: cross Abstract: Extending a vision-language-action (VLA) policy to a new task typically requires task-specific teleoperated demonstrations and per-task fine-tuning, making adaptation costly in both data collection and compute.
By Jeongeun Park, Juhan Park, Taekyung Kim, Sungjoon Choi, Dongyoon Han, Sangdoo Yun
arXiv:2603.16065v3 Announce Type: replace-cross
Abstract: Reinforcement Learning (RL) has shown strong potential for improving robotic manipulation policies, yet its practical use remains bottlenecke...
By Yanru Wu, Weiduo Yuan, Esteban Martinez Licon, Ang Qi, Vitor Guizilini, Jiageng Mao, Yue Wang
The paper introduces D&A (Describe & Assess), a prompt-based method that separates progress estimation into a textual description of scene dynamics and a subsequent assessment using explicit rubrics. By requiring Vision‑Language Models to generate an explicit definition of absolute success before scoring, D&A anchors evaluations to an external criterion, reducing confirmation bias and providing fine‑grained, partial progress metrics. Empirical results show that D&A matches or surpasses post‑trained reward models across tasks, embodiments, and benchmarks, and its reward estimates support efficient downstream reinforcement learning.
By Marius Memmel, Shawn Yin, Urvi Rutia, Angad Wadhwa, Eldon Li, Gavin Angeloff, Ankit Goyal, Jesse Zhang, Dieter Fox, Fabio Ramos, Anqi Li, Abhishek Gupta
Action-conditioned latent world models predict future visual representations, enabling zero-shot goal-conditioned robot planning and control. However, their predictions for fine-grained spatial and ro...
The paper introduces D&A (Describe & Assess), a prompt‑based method that separates vision‑language model progress estimation into a textual description of scene dynamics and a rubric‑based assessment. By requiring the model to explicitly define absolute success before scoring, D&A reduces confirmation bias and provides fine‑grained, partial progress metrics that accurately penalize failures and interpolate successes. The approach generalizes across tasks, embodiments, and benchmarks, matching or surpassing post‑trained reward models while enabling efficient downstream reinforcement learning.