Why Self-Improving Autoresearch Loops Fail: Paper Reveals Sharp Drop in LLM Judge Accuracy
dair_ai · x · 2026-08-05
A new study reveals why self-improving autoresearch loops become brittle. The research shows that an LLM's accuracy in judging its own ideas degrades sharply as iterations increase.
In public AutoSOTA logs, the fraction of helpful modifications falls from 70% in the first two iterations to 43% by iteration six. Across a 366-pair benchmark, as successful changes accumulate, selective accuracy plummets from 82.8% to 56.9%, yet the LLM judge remains just as willing to make decisions. The paper suggests small loop changes like "Rehearse" to mitigate this issue.
More from AGI Musings
- Rumor: SSI Solved Continual Learning, But Alignment Becomes Harder — teortaxesTex · 2026-08-05
- Experts from Odyssey and 1x to Discuss World Models in Embodied AI — Exp_Mark · 2026-08-05
- Top Researchers Launch 'Pax Machina' to Debate Institutions for Powerful AI — edelwax · 2026-08-05
- Settled Science vs. Breaking News: Why Core Psychology Findings Hold Up — Astral Codex Ten · 2026-08-05
- Restructuring Work in the AI Era: Become the Designer of Your Own Job — every · 2026-08-05
- Musk: AI and Robotics Will Trigger Unprecedented Global Bandwidth Demand — XFreeze · 2026-08-05