BWhat changed
Trained the vision-language backbone with discrete-action/text prediction while separately training a continuous flow-action expert, stopping the expert's gradients into the backbone to preserve useful representations.
Scientific scope
Trained the vision-language backbone with discrete-action/text prediction while separately training a continuous flow-action expert, stopping the expert's gradients into the backbone to preserve useful representations. Stop-gradient, discrete action tokens and flow-action experts are inherited. Insulation is a specific training recipe, not guaranteed preservation of every pretrained fact or capability. pi0/pi0.5 system records do not separately own this original method.
Evidence
The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. Stop-gradient, discrete action tokens and flow-action experts are inherited. Insulation is a specific training recipe, not guaranteed preservation of every pretrained fact or capability. pi0/pi0.5 system records do not separately own this original method.
Contributors
- Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.
- Driess et al.: gradient-insulated continuous-action experts — original technical report
- Driess et al.: gradient-insulated continuous-action experts — institutional account
Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.
Full scientific record · Browse contribution grades · Request a correction