MLE-World Learns From Past ML Experiment Traces, Gaining 19% on MLE-bench Lite
sherryyangML · x · 2026-10-23
MLE-World turns execution traces from past experiments (code + task context → errors, logs, scores) into a predictive model, letting agents revise ideas before spending real compute.
On 17 MLE-bench Lite tasks:
- Mean leaderboard percentile: 45.4 → 54.1, +19% over baseline
- Real execution time: 9.94 → 4.77 hours per trial, less than half
Agents use the learned feedback to refine ideas and only run real experiments when predictions are uncertain.
More from coding & agent
- Developers Don't Code Anymore — Now Their Job Is Rotating Secrets Daily — Dan_Jeffries1 · 2026-10-23
- New paper: coding agents in a shared sandbox self-create rules, reciprocity debts, and coalitions — nandofioretto · 2026-10-23
- Developer runs AI project managers that manage their own AI agents to build his game — draginol · 2026-10-23
- RentAKaren: a real service that rents humans to handle complaints your AI agent can't win — mathemagic1an · 2026-10-23
- Clerk's Auth for MCP growing 13% week-over-week, fastest since 2023 — stuffyokodraws · 2026-10-23
- Shopify builds remote mobile dev environments for AI agents — rseroter · 2026-10-23