World Models Dream of Success: Diagnosing and Repairing Failure Insensitivity in Robot World Models

Key points

• Across four released checkpoints from two architecture families, we observe weak sensitivity to action changes and success-like predictions on verified failures.1
• Although recent work incorporates failures into model training, which data can repair released checkpoints without changing their architecture or training objective still remains underexplored.1
• To this end, we introduce CureWM, which constructs alternative actions from successful demonstrations across a severity grid, verifies their outcomes through execution in simulation or on hardware, and fine-tunes released models on the resulting failures and surviving successes alongside nominal demonstrations.1
• These findings support execution-verified counterfactual replay for post-hoc repair and show why reduced optimism must be evaluated alongside success--failure discrimination.1

Sources (1)

• [1]World Models Dream of Success: Diagnosing and Repairing Failure Insensitivity in Robot World Models
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 09:26 PM

  Robot world models support policy evaluation, planning, and synthetic data generation, but these applications require predictions that distinguish successful actions from failures.

  Across four released checkpoints from two architecture families, we observe weak sensitivity to action changes and success-like predictions on verified failures.

Extractive summary: sentences quoted from the sources.