PyRUA-Lean Boosts GPT-6 Astra Robot Agents 14% Success with 65% Fewer Tokens
DAGroup-PKU · hf · 2026-10-02
PKU's DA Group introduces PyRUA-Lean, an interactive code-execution framework that slashes token overhead for VLM-based robot agents:
- Feedback-driven primitive composition: the agent composes classical robot primitives and learned VLA policies into Python cells with conditional checks and local retries, returning only explicitly requested images and state feedback for replanning.
- Results: across 700 simulated tasks from LIBERO-PRO, RoboTwin 2.0, and RoboCasa365, using the same GPT-6 Astra planner, overall success rises from 63.1% to 71.7% under equal LLM-call budgets; on instances both agents solve, it uses 49% fewer LLM calls and 65% fewer input tokens.
More from Embodied
- Dutch greenhouse deploys AI-powered FANUC robot arm to pick ripe tomatoes — anselm · 2026-10-02
- EgoTools: 100-Hour Egocentric Dataset Shows Video Models Struggle at Tool Reasoning — ropedia-ai · 2026-10-02
- Neuralink pretrains decoders on 50,000+ hours of neural data—dataset may be the real moat — CurieuxExplorer · 2026-10-02
- Construction Is 13% of Global GDP: Gravis Retrofitting Existing Machines for Autonomy — lukas_m_ziegler · 2026-10-02
- Saga Robotics' Thorvald fleet kills vineyard mildew with night-time UV-C, no chemicals — TinfoilTricorn · 2026-10-02
- High schoolers built a ball-collecting robot in 6 weeks using same mechanics as $2,995 Tennibot — TinfoilTricorn · 2026-10-02