Frontis-MA1: Training an AI4AI Model towards Recursive Self-Improvement in Machine Learning Engineering

Horizon Research, Frontis.AI, Tsinghua University

Research official 1 src. ~1 min

Frontis-MA1 (35B) is a meta-evolution agent post-trained around four program-evolution operators (Draft, Improve, Debug, Crossover) inside OpenMLE, an open full-stack recursive self-improvement testbed for machine learning engineering with verifiable execution feedback.

Why it matters

Gathered 165 upvotes on HuggingFace Daily Papers; on MLE-Bench Lite it lifts Medal Average from 39.39% to 60.61% (71.21% with extended search) over its base model, with full model weights and stack released.

Importance: 3/5

Notable paper with 165 HuggingFace Daily Papers upvotes (>=100 threshold bump applied).

Sources