Meta's A-MLE agent automates ML experimentation for ads ranking, cutting error 2.56%
rohanpaul_ai · x · 2026-09-22
Meta has built and deployed A-MLE (Agentic ML Exploration), an autonomous LLM-agent system that automates the ML iteration loop for its ads-ranking models: hypothesis generation, experiment execution, failed-job recovery, result comparison, and cross-model knowledge sharing.
The motivation: production ML is now bottlenecked by human iteration throughput—each statistically significant improvement takes days to weeks of senior engineer time per model, and techniques diffuse slowly across heterogeneous ranking stacks. Key numbers:
- A generic LLM scored 8% on Meta's basic capability test; the domain-equipped agent reached 68%
- On one experimental ranking model, broader exploration cut offline regression error by 2.56% relative to baseline while training QPS stayed essentially unchanged (+0.42%)
The paper is on arXiv (2609.08248) with 39 authors.
Related event: Meta Open-Sources A-MLE: LLM Agents Automate Ad Ranking ML Experiments(2 posts)→
More from coding & agent
- Pigment Agents now reads and posts in Slack, inheriting user permissions — mattturck · 2026-11-12
- Map of the agent economy: 50+ startups across payments, spend control, identity, memory — venturetwins · 2026-11-12
- OpenCode launches Browse, an on-device browser agent with local logins — Saul_Loveman · 2026-11-12
- VentureTwins maps the B2A economy: dozens of agent infra and app companies in first series installment — venturetwins · 2026-11-12
- Google releases 76-page 'Agents Companion' whitepaper on agent engineering — mdancho84 · 2026-11-12
- Spinning up containers for 20-line agent tools is backwards: Safescript compiles provable, 50ms-deploy apps — uriwa · 2026-11-12