Skip to content

Event

RoboArchitect: Automatically Designing Robot Harnesses Without Updating Weights

1 reports · 1 sources

Shengguang Wu and other authors released a paper proposing RoboArchitect: letting coding models observe robot rollouts, diagnose failures, and iteratively modify the agent framework (harness) surrounding a frozen policy, while evolving persistent memory. The authors claim that, without updating any model weights, this method raises the average success rate on unseen scenarios from 9.7% to 43.3% (4.5x) across multiple benchmarks and policies, and on unseen tasks from 12.7% to 40.2% (3.2x). The harness is evolved and optimized over a set of episodes, and evaluated on held-out scenarios and tasks. Based on this, the authors argue that harness design can be automated, and is an alternative to fine-tuning under the new setting.

Report timeline

  1. RoboArchitect: Automating the Design of Robotic Agents with Persistent Memory

    RoboArchitect lets a coding model observe robot rollouts, diagnose failures, and iteratively revise the agent harness around a frozen policy. Without updating any model weights, it raises average success from 9.7% to 43.3% (4.5x) on unseen scenes and from 12.7% to 40.2% (3.2x) on unseen tasks.

    arXiv cs.RO ·

Related topics

Grasping and manipulationRobot models and learningResearch 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.