Fewer Tokens, Better Action: GPT-6 Astra Robot Agents with 14% Higher Success Rate but 65% Fewer Tokens
PyRUA-Lean beat the same GPT-6 Astra planner’s tool-calling baseline while spending far less context on repeated robot observations.
The paper tests an interactive code-execution setup that lets the robot agent compose primitives, run conditional checks, and retry locally before asking the model to replan. Across 700 simulated tasks, success rose from 63.1% to 71.7% under the same LLM-call budget. On tasks both agents solved, PyRUA-Lean used 49% fewer LLM calls and 65% fewer input tokens. HF Daily Papers' note
The paper tests an interactive code-execution setup that lets the robot agent compose primitives, run conditional checks, and retry locally before asking the model to replan. Across 700 simulated tasks, success rose from 63.1% to 71.7% under the same LLM-call budget. On tasks both agents solved, PyRUA-Lean used 49% fewer LLM calls and 65% fewer input tokens. HF Daily Papers' note
score 5