Agent That Rewrites Optimizer Code on the Fly Matches Best Hand-Tuned Recipes
ChengleiSi · x · 2026-10-22
Researchers propose a "reactive optimizer": an agent watches training dynamics and rewrites the optimizer's source code while the network trains, letting optimization decisions adapt in real time.
Across language and image settings without a tuned recipe, it matches or beats the best one-shot recipe, and it autonomously guards against loss spikes and overfitting. The authors argue optimization decisions may soon belong to agents rather than humans. Paper and runnable code are available.
More from coding & agent
- Dev uses 'Claude Tag' to auto-handle a pile of open PRs without babysitting each one — maxleiter · 2026-10-22
- Agent rewrites optimizer source code mid-training to build a reactive optimizer — giffmana · 2026-10-22
- GMG Open-Sources an Apache 2.0 ServiceNow Agent With Human-Approved Changes — adnan_hashmi · 2026-10-22
- RAX drives robots with a frozen coding agent; DeepSeek version beats π0.5 at 1/20 cost — du_yilun · 2026-10-22
- Stripe partners with Anthropic to run business payments and refunds inside Claude Duet — jeff_weinstein · 2026-10-22
- Aaron Levie: autonomous domain-specific agents are AI's emerging form factor, a leap past the LLM-wrapper era — _AustinCalvert_ · 2026-10-22