Skip to content

Event

TinyAstral: Self-Evolving Harness Transfer to a 4B Manipulation Agent

1 reports · 1 sources

Kehong Gong and co-authors submitted the paper TinyAstral, proposing a two-stage method: first freeze an open-weight teacher model and use recursive self-improvement (RSI) to optimize its harness, with each modification accepted only if it passes a paired statistical test; then transfer the optimized harness to a distilled 4B student model. On RoboTwin, harness evolution raised the teacher's success rate from 33.4% to 63.6%, and the final 4B student reached 62.8%; in measured deployment episodes, the student used about one-fifth of the teacher's accelerator time. These figures come from the paper's own account and have not yet been independently verified.

Report timeline

  1. TinyAstral: Transferring a self-evolving harness to small generalist manipulation agents

    TinyAstral proposes a two-stage method: first freeze an open-weight teacher model and optimize its harness through recursive self-improvement (RSI), then distill a 4B student model to replace the teacher within the same harness. On RoboTwin, harness evolution raised the teacher's success rate from 33.4% to 63.6%, and the final 4B student reached 62.8%; in measured deployment episodes, the student used about one-fifth of the teacher's accelerator time.

    arXiv cs.RO ·

Related topics

Grasping and manipulationRobot models and learningData and simulationResearch papers

Interest trend

Current
3
Peak
4Nov 4 01:00
Past 24 hours
–

The trend compares participants observed throughout the same period, so its range may be smaller than the current interest count. Move or tap to inspect each hour; use the arrow keys on a keyboard.