- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 3
- Soundness
- 2
Read review
Paper summary
This paper introduces DELE-w0.5, a robot policy jointly trained to predict action chunks and future visual features using flow matching. A structured attention mask prevents action tokens from attending to future-state tokens, allowing the future branch to be removed at inference. Evaluated against seven baselines on four real-world manipulation tasks, the method achieves 62.5% overall task success and 81.3% average ordered-stage progress.
Critical strengths
- The method achieves 62.5% full-task success across four multi-stage manipulation tasks, outperforming the strongest evaluated baseline by 32.5 percentage points.
- The attention mask allows future-state tokens to be removed at inference while retaining their supervision during training.
- Reporting both ordered-stage progress and full-task success distinguishes partial execution from completion. The tasks include challenging bimanual coordination and articulated-object manipulation.
Critical weaknesses
-
Prior work such as LaDi-WM and FLARE already explores future latent prediction for robot control. The paper does not clearly establish what its joint flow-matching objective and attention mask add beyond these approaches.
-
There is no controlled comparison against the same policy trained without future-state supervision. The reported gains therefore cannot be attributed to the proposed mechanism rather than pretraining data or other design choices.
-
Action tokens cannot attend to future-state tokens, which are removed at inference. This supports future prediction as auxiliary supervision, but does not directly demonstrate “inferring action from future latent state.”
Questions for authors
-
How much does performance drop when future-state supervision is removed under otherwise identical training conditions?
-
Does flow matching for future features outperform a simple feature-regression loss? What benefit does the proposed attention mask provide?
-
What distinguishes the method from FLARE-style future representation alignment, and can this difference be tested under matched conditions?
-
Given that action tokens cannot access future-state tokens, what precisely does “inferring action from future latent state” mean in this architecture?
-
Does the method retain its advantage under randomized object positions or systematic execution-time perturbations?