WCM: A JEPA-Based World Critic Model Beats SOTA on 149 VLA Robot Tasks
jiqizhixin · x · 2026-08-22
Robots keep failing at memory-dependent tasks because their critics only look at a single frame, missing temporal dynamics. The World Critic Model (WCM) from Fudan University, Shanghai Innovation Institute, and Tongji University fixes this with a lightweight JEPA-based architecture that predicts future latent states and estimates values — the explicit world-modeling objective forces the critic to learn cause-and-effect over time, not just scalar rewards.
WCM plugs into top VLA backbones (π0, π0.5, OpenVLA-OFT) and beats state-of-the-art on 149 tasks across four benchmarks, with especially strong out-of-distribution generalization. Seven real-world manipulation tasks confirm it's not just a simulation trick. Paper, code, and dataset are all released.
More from Embodied
- Tevel's robotic system harvests fruit around the clock using computer vision — LexiLove · 2026-10-16
- SHOES: 10 minutes of real-robot RL lets quadrupeds walk in shoes, CoRL 2026 paper — kscottz · 2026-10-16
- Tesla Cybercab robotaxis logged zero crashes in first two weeks of public rides, NHTSA data shows — elonmusk · 2026-10-16
- Robotics LLMs tested on 16 tasks: GPT-6 Astra hits 17.2% success vs Fable's 9.1% — pabbeel · 2026-10-16
- WEDGE paper generates training environments conditioned on navigation agent failures — _akhaliq · 2026-10-16
- Viral meme nails AI's irony: it can write symphonies but can't do the dishes — briannekimmel · 2026-10-16