AI
Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation
Publication: Hugging Face Daily Papers
Author: Not available in collected metadata
Source published: 28/09/2026, 20:00:00 UTC
Briefing created: 01/10/2026, 09:22:02 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 01/10/2026, 01:18:24 UTC.
Briefings
Researchers propose PRISM, a real-to-sim-to-real framework that generates hundreds of diverse counterfactual human-object interaction videos from a few exemplar clips to enable scalable humanoid locomotion. By amplifying limited real-world data, the system trains a single policy that generalizes to unseen objects and initial configurations without fine-tuning, successfully demonstrating object pickup, carrying, and dropping across novel instances using only onboard depth observations.
Key points
- The PRISM framework uses video-to-video generation to create hundreds of diverse counterfactual videos from a small number of exemplar real videos.
- A contact-anchored real-to-sim pipeline reconstructs human and object motions to produce physically plausible trajectories for training a generalizable policy.
Context and limitations
The system operates without real-world fine-tuning and relies solely on onboard depth observations for navigation and interaction.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 01/10/2026, 09:22:02 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeidndsxyoecg43qzq4hn5iz4cjellrecyk7imeq7opjp64evgvgkpi