Microsoft proposes Harnessed Agentic RL to bridge training-deployment gap
SergioPaniego · x · 2026-08-20
Microsoft Research released Agent Lightning v1.0 and a paper proposing the "Harnessed Agentic RL" paradigm. This approach advocates training agents directly within their real-world deployment harnesses (which manage tools and context) rather than in reimplemented environments, thereby narrowing the gap between training and actual use. The paper details challenges unique to this method, such as retokenization, sample merging, advantage calculation, loss normalization, and backend scheduling, offering corresponding solutions.
Related event: Microsoft Releases Agent Lightning v1.0, Boosting SWE-bench to 56.4%(3 posts)→
More from coding & agent
- Claude Made a Styled Video in Four Steps—and Caught a One-Pixel Flaw Itself — aakashgupta · 2026-10-16
- Datadog Launches Scout 1.0, an RL-Trained Model for Investigating Production Incidents — HanchungLee · 2026-10-16
- Microsoft paper: history-aware model routing lifts multi-turn agent task pass rate by 3.3 points — dair_ai · 2026-10-16
- ORA: open robot-use agents powered by DeepSeek V4.1-Flash compete with frontier models — JunGao33210520 · 2026-10-16
- Non-developer ships a full platform fighter in 30 days with AI agents — Teknium · 2026-10-16
- 2 humans + 5 agents: plain-English feature requests now drive their codebase autonomously — nptacek · 2026-10-16