Research map / Reinforcement learning / Visual and representation RL
ChronoFlow Policy: Unifying Past-Future Interaction Flow in Visuomotor Policy Learning
Bokai Lin, Yifu Xu, Xinyu Zhan, Hongjie Fang, Jialin Tian, Fu-Cheng Zhang, Yong-Lu Li, Cewu Lu, Lixin Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add historical ChronoFlow tokens (ablation) | CFP w/o past · Real-world Swap-Easy stage III | success rate (%): 20 -> 93 | Improved |
| add historical ChronoFlow tokens (ablation) | CFP w/o past · Real-world Swap-Hard stage III | success rate (%): 11 -> 61 | Improved |
| add historical ChronoFlow tokens (ablation) | CFP w/o past · Real-world Prepare Breakfast stage II | success rate (%): 67 -> 80 | Improved |
| add ChronoFlow future supervision (interaction-centric flow) | DP3 · MetaWorld (7 tasks) | avg success rate (%): 30 -> 72 | Improved |
| use interaction-centric ChronoFlow instead of dense scene flow | 3D-FDP · MetaWorld (7 tasks) | avg success rate (%): 34 -> 72 | Improved |
| use ChronoFlow instead of object-pose trajectories | MBA · RoboTwin 2.0 (7 tasks) | avg success rate (%): 56 -> 66 | Improved |
| use Unet1D instead of DiT backbone | CFP · MetaWorld (7 tasks) | avg success rate (%): 70 -> 72 | No clear effect |
| add ChronoFlow trajectory supervision (ablation) | CFP · Real-world Fold Towel stage II | success rate (%): 67 -> 87 | Improved |
| add object flows to ChronoFlow (ablation) | CFP · Real-world Fold Towel stage II | success rate (%): 47 -> 87 | Improved |
| add gripper flows to ChronoFlow (ablation) | CFP · Real-world Fold Towel stage II | success rate (%): 80 -> 87 | Improved |
| add stochastic keypoint subsampling during training (ablation) | CFP · Real-world Fold Towel stage II | success rate (%): 27 -> 87 | Improved |
| add ChronoFlow spatial jitter augmentation (ablation) | CFP · Real-world Fold Towel stage II | success rate (%): 6 -> 87 | Improved |
Closest papers by meaning
- Latent Policy Barrier: Learning Robust Visuomotor Policies by Staying In-Distribution (NeurIPS 2025 · similarity 0.90)
- FreqPolicy: Frequency Autoregressive Visuomotor Policy with Continuous Tokens (NeurIPS 2025 · similarity 0.90)
- Chain-of-Action: Trajectory Autoregressive Modeling for Robotic Manipulation (NeurIPS 2025 · similarity 0.88)
- FreqPolicy: Efficient Flow-based Visuomotor Policy via Frequency Consistency (NeurIPS 2025 · similarity 0.88)
- DynaMo: In-Domain Dynamics Pretraining for Visuo-Motor Control (NeurIPS 2024 · similarity 0.87)
- Falcon: Fast Visuomotor Policies via Partial Denoising (ICML 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
