RobotAPO: Adversarial Physics Preference Optimization for Robotic Manipulation Video Generation

Key points

• Because standard supervised fine-tuning lacks the direct pressure to penalize these localized failures, we introduce AgiBot-PhysPref.1
• Building upon this, we propose RobotAPO, an adversarial physics preference optimization framework operating in the continuous flow-matching denoising space.1
• To prevent the policy from merely memorizing static curated failures, RobotAPO employs a lightweight adversarial counterfactual proposer that learns a condition-dependent, physical-failure-biased direction in denoising space.1
• Comprehensive evaluations demonstrate that explicitly correcting these localized physics violations improves downstream robot execution from generated videos.1

Sources (1)

• [1]RobotAPO: Adversarial Physics Preference Optimization for Robotic Manipulation Video Generation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 05:10 AM

  Robotic manipulation videos are increasingly used as visual plans for embodied agents, but optimizing purely for visual plausibility often fails to capture the fragile physical manifold of real-world interactions.

  Because standard supervised fine-tuning lacks the direct pressure to penalize these localized failures, we introduce AgiBot-PhysPref.

Extractive summary: sentences quoted from the sources.