Custom MLE-Agent hits 100% gold on MLE-Bench Tiny on a 2-vCPU EC2 box

InfamousTrouble7993 · reddit · 2026-11-20

A developer unveiled a self-built MLE-Agent scoring 100% gold medals on MLE-Bench Tiny and 60% medals on MLE-Bench Small, with an approach claimed to differ significantly from state-of-the-art agents — running on a modest 2 vCPU / 4 GB RAM EC2 instance. It handles tabular, vision, audio, sound and science modalities (no generation yet), with a paper planned. The author is seeking partners for business case studies.

Original post →

More from coding & agent

coding & agent channel →