Reka RIDM: game-trained motion labels for real-video world-model data

What changed

Released compact game-trained inverse-motion models, comparing optical-flow inputs with direct pixels and using filtered/resampled game controls to infer movement/camera labels for unlabeled videos.

Scientific scope

Released compact game-trained inverse-motion models, comparing optical-flow inputs with direct pixels and using filtered/resampled game controls to infer movement/camera labels for unlabeled videos. Inverse dynamics, RAFT optical flow, IMPALA and VPT-style encoders are inherited. Inferred game-key/camera labels on video are not measured robot joint torques or true causal motor commands. Real-video evaluations are small; some metrics lack run logs and are marked unconfirmed. The pixel model explicitly fails on real-world footage. Code/weights license does not license evaluation media or game data.

Evidence

The linked original sources describe this contribution and its evidence limits. WRS has not independently reconstructed the full scientific claims. Inverse dynamics, RAFT optical flow, IMPALA and VPT-style encoders are inherited. Inferred game-key/camera labels on video are not measured robot joint torques or true causal motor commands. Real-video evaluations are small; some metrics lack run logs and are marked unconfirmed. The pixel model explicitly fails on real-world footage. Code/weights license does not license evaluation media or game data.

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.

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