Frontis-MA1 Paper: Training an AI4AI Model for Recursive Self-Improvement
mark_k · x · 2026-08-02
This post shares the paper details for Frontis-MA1. The paper explains that Recursive Self-Improvement (RSI) requires systems that improve the process of building AI (AI4AI), with Machine Learning Engineering (MLE) serving as an executable testbed.
The researchers introduced OpenMLE, an open full-stack system covering task environments, operator learning, and long-horizon search. The trained 35B agent utilizes execution feedback and asynchronous search to hit 71.21% on MLE-Bench Lite, beating GPT-5.5 + Codex and approaching Kimi K3. On the held-out NatureBench Lite, swapping in this model raises Match-SOTA from 50% to 70%, proving its generalization and self-evolution capabilities.
Related event: OpenMLE: An Open-Source Stack for Recursive AI Self-Improvement(3 posts)→
More from coding & agent
- Connected steps are not a workflow: nobody designs the handoff — usehive · 2026-10-16
- Making Blender animations with Opus, Astra and Argon: how you prompt still matters — chongdashu · 2026-10-16
- Running PolyStrata + Cline coding agent on a $0-fancy laptop at ~90 tok/s — rpwoerk · 2026-10-16
- How do you attribute agent gains: model training or infrastructure? — jrxna · 2026-10-16
- Feed Claude your books, blogs and top engineers' code for claude.md — 'it now writes better code than me' — blaizedsouza · 2026-10-16
- Open-source /label plugin auto-names and colors Claude Code sessions in one step — hermanschutte · 2026-10-16