LeCun coauthors prove standard world models fail at counterfactuals, fix needs no retraining
alex_verem · x · 2026-11-22
Yann LeCun, Randall Balestriero and Kerda Varaku (Brown/NYU, AMI Labs) prove that world models trained on factual transitions cannot answer counterfactual questions — what would have happened under a different action given the same hidden disturbance — regardless of data or capacity. Their one-line fix: counterfactual outcome ≈ observed outcome plus the model's predicted difference between the two actions, exploiting that action–disturbance interactions are second-order in time. No counterfactual data, extra parameters, or retraining needed. They prove an O(h²) error bound and a matching minimax lower bound, characterize failure cases (discontinuities, coordinate folds, command-scaling disturbances), and validate on JEPA world models over frozen DINOv2/MAE encoders across four rendered simulators (robot hand, arm, ball-in-cup, liquid carrying), lifting counterfactual R² from 38–46% to 86–96%.
More from Embodied
- Heavy-duty robot shows off gym-bro strength in new demo — LinusEkenstam · 2026-11-22
- Slush anecdote: 50 people pitched data/ML platforms, zero sold robotics hardware — MarvinTBaumann · 2026-11-22
- Caltech's ASTRA robot autonomously builds 1.4m antenna in orbit, a space first — mattbeane · 2026-11-22
- Robotics won't wake up to a 250-page PDF declaring it 'solved,' quips roboticist — DJiafei · 2026-11-22
- Robotics data companies capture nearly 50% of attention so far — chris_j_paxton · 2026-11-22
- PredActor Cuts Diffusion to 5% of Forward Pass for Low-Latency Humanoid Control — chris_j_paxton · 2026-11-22