MIT CSAIL's Workspace Models help robots learn complex skills with millisecond-level planning
rishabh16_ · x · 2026-11-12
MIT CSAIL and CMU's MLD introduced "Workspace Models," a new memory approach covered by MIT News.
Problem: Robots err on home and factory chores because they struggle to extract relevant details from past observations in real time; VLM-based agent pipelines that search long frame histories at runtime cause errors, high compute costs, and poor reactivity.
Approach: Large VLMs reason offline to train a smaller module that converts observation history into latent vectors encoding action-relevant information, using one VLM to review videos and another to extract spatial data.
Result: Robotic arms can plan their next move in milliseconds in response to environment changes.
More from Embodied
- Stanford's FIDO robot dog learns physics by playing fetch with humans — kevin_zakka · 2026-11-12
- EqC paper: introspective robot policies that know when to ask for help — du_yilun · 2026-11-12
- iPhone Duo fold test hits record 413,000 cycles in 13 days with zero failures — SumitGup · 2026-11-12
- kevin_zakka adds tactile sensing support to mjlab robotics framework — kevin_zakka · 2026-11-12
- MCP Display turns an old Android phone into a programmable screen for AI agents — PatrickCavalcanti · 2026-11-12
- Dennis Hong showcases THOR hazard-response robot and ARTEMIS at AIVEX Frontier 2026 — DennisHongRobot · 2026-11-12