Train lightweight world action models for robot manipulation using state-fusion decoding and frozen video backbones on LIBERO and RoboTwin2.0.
microsoft python machine-learning websocket parallel pubsub distributed knowledge-graph webapp gbm lightgbm gpt pub-sub wamp-protocol websocket-subprotocol remote-procedure-calls large-language-models genai graphrag vision-language-action-model
-
Updated
Jun 28, 2026 - Python