Paper 2607.28568
Frontis-MA1: Training an AI4AI Model towards Recursive Self-Improvement in Machine Learning Engineering
- Published
- Jul 2026
- Research lab
- Independent
- Citations
- 0
- GitHub
- 212 stars
01 In brief
Summary
The paper introduces OpenMLE, an open full-stack system for studying recursive self-improvement (RSI) in machine learning engineering (MLE), and Frontis-MA1-35B, a meta-evolution agent trained on this stack.
OpenMLE comprises OpenMLE-Gym (5,758 quality-gated executable tasks with sandboxed execution), OpenMLE-ERL (execution-grounded SFT and RL training of four atomic operators: Draft, Improve, Debug, Crossover), and OpenMLE-Evo (experience-guided long-horizon search with multi-factor parent selection and operator-conditioned memory).
On MLE-Bench Lite under a 12-hour per-task budget on one RTX 4090 (12 GB VRAM), Frontis-MA1-35B improves Medal Average from 39.39% (base Qwen3.6-35B-A3B) to 60.61% with OpenMLE-Evo, and to 71.21% with OpenMLE-Evo-Max, exceeding GPT-5.5 + Codex and approaching GPT-5.6 Sol and Kimi K3.
On held-out NatureBench Lite, the trained model raises Match-SOTA from 50% to 70% with the framework fixed, and OpenMLE-Evo raises it from 20% to 50% with the model fixed.
The authors release model weights and the full OpenMLE stack for reproducible research on executable AI4AI toward RSI.
02 From the paper
Abstract
Recursive self-improvement (RSI) requires AI systems that improve the process of building AI (i.e., AI4AI); machine learning engineering (MLE) offers a concrete, executable testbed for studying this capability. We introduce OpenMLE, an open full-stack system for RSI research in MLE, spanning verifiable task environments with execution feedback (OpenMLE-Gym), operator learning (OpenMLE-RL), and long-horizon search (OpenMLE-Evo). On this stack we post-train Frontis-MA1 (35B) as a meta-evolution agent for MLE, aligning post-training and inference around four atomic program-evolution operators (Draft, Improve, Debug, Crossover): the same operators are trained via execution-grounded SFT and RL on data deduplicated against all evaluation benchmarks, then composed into long-horizon search, coupling learning and evolution in a single loop. On MLE-Bench Lite under a 12-hour per-task budget on one RTX 4090 capped at 12 GB VRAM, Frontis-MA1 (35B) improves Medal Average from 39.39% to 60.61% over its base model with OpenMLE-Evo, and reaches 71.21% with OpenMLE-Evo-Max (benchmark-independent experience priors and asynchronous search), exceeding GPT-5.5 + Codex and approaching GPT-5.6 Sol and the 2.8T Kimi K3. On held-out NatureBench Lite, both components transfer: with the framework fixed, swapping in the trained model raises Match-SOTA from 50% to 70%; with the model fixed, swapping in OpenMLE-Evo raises it from 20% to 50%. We release the model weights and the full OpenMLE stack to enable reproducible research on executable AI4AI toward RSI. Code: https://github.com/FrontisAI/OpenRSI