Code & Chain · Signal Desk

Microsoft Open-Sources Orchard Framework for Scalable AI Agent Training and Evaluation

Original sourceMicrosoft SignalAdditional: ETIH EdTech News

Summary

Microsoft Research releases Orchard, an open-source framework for training and evaluating AI agents across software engineering, web navigation, and personal assistant tasks. Orchard offers Orchard Env, a Kubernetes-native service, to manage data collection, reinforcement learning rollbacks, and ev…

Key points

  • Developers and researchers can use Orchard to build their own agent training pipelines and integrate with tools like Codex.
  • Microsoft's open-source agent training framework lowers the barrier for AI agent development and demonstrates performance gains with Codex integration.
  • Developers can train custom agents and enhance existing tools like Codex, accelerating agent deployment.

Editorial note

This page is Code & Chain's editorial summary of public sources. It may be prepared with AI assistance and published through an automated workflow. Refer to the original sources; this content is not investment, legal, or tax advice.