DreamTrue predicts future observations across camera views and robot embodiments.
Counterfactual post-training improves robot world model's action following and physical plausibility
DreamTrue combines offline geometric calibration with reward-guided reinforcement learning on human-annotated defect videos, cutting interaction defect rates from 48.12% to 6.25% on AgiBot.
Chinese Tech
Junyan Li · Ruizhi Li · Yu Liu · Xiangshuo Liu · Mingchao Sun · Hongyu Pan · +3 more
Institute of Automation, Chinese Academy of Sciences · Alibaba Group
Research Digest··2 min read
The authors present DreamTrue, a multi-view robot world model that predicts future video frames conditioned on action sequences.
Why this paper
From Institute of Automation, Chinese Academy of Sciences and Alibaba Group
In one line
A robot world model using offline geometric calibration and counterfactual post-training improves action following and cuts human-assessed interaction defects from 48.12% to 6.25%.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§