Today for AI
HOT RADAR
agentHEAT 15.2°

AI CURATED HOT STORY · 10/23/2026

SenseTime Releases RoboRSI-27B Distilling GPT-6-Astra Embodied Experience at 1/50 Cost

1 reports archived1 independent sourcesupdated 10/23/2026, 23:32:59
Synthesis & Latest Updates
1 Sources Cross-Validated

SenseTime open-sourced the SenseNova-RoboRSI framework and a 27B model, distilling GPT-6-Astra's embodied intelligence via 'experience internalization.' This approach maintains active perception, spatial reasoning, and planning capabilities while reducing task execution costs to approximately 1/50 of the original model, highlighting current bottlenecks in real-time responsiveness and economic feasibility for embodied AI.

LATEST/SenseTime open-sourced the SenseNova-RoboRSI framework and a 27B model, distilling GPT-6-Astra's embodied intelligence via 'experience internalization.' This approach maintains active perception, spatial reasoning, and planning capabilities while reducing task execution costs to approximately 1/50 of the original model, highlighting current bottlenecks in real-time responsiveness and economic feasibility for embodied AI.

HEAT TRENDHourly heat curve

2 fully observed hours
Fewer than 3 fully observed hours; trend not plotted yet.

TIMELINECoverage timeline

Total 1 reports · Latest first
  1. Rohan Paul (@rohanpaul_ai)T2·88 pts
    • Core mechanism is 'experience internalization,' using ICL to extract visual, state, and trajectory info from demonstrations and distill it into smaller models
    • RoboRSI-27B demonstrates active perception, spatial reasoning, and feedback-driven revision across simulation and real-robot benchmarks
    • While GPT-6-Astra sets new SOTAs under unified protocols, high inference costs and slow response times lead to failures in physical world tasks