AIDE² paper: self-improving research agent beats the version humans tuned for two years
egrefen · x · 2026-09-24
The arXiv paper on AIDE² is out. It's an RSI (recursive self-improvement) system where AI research agents improve their own research efficiency.
Key result: AIDE² produced a research agent that beats the version the team hand-tuned for two years, and the gains hold on benchmarks the outer loop never saw.
New in the paper: results on transfer across models and comparisons with more AI research agents.
Related event: AIDE²: Self-Improving Research Agent Beats Two Years of Human Tuning(3 posts)→
More from AGI Musings
- "Worry about the dumbest thing a model can not do," says MillionInt — MillionInt · 2026-09-24
- Health organizations use Claude to fight rare Ebola outbreak in DR Congo — ResultBackground2450 · 2026-09-24
- Gary Marcus: Musk quietly pushed his AGI prediction from 2025 to 2028, same pattern as FSD — GaryMarcus · 2026-09-24
- Raising VC will feel like a sign of inefficiency as AI lets companies build core tech in-house — eherrerosj · 2026-09-24
- Bengio revisited: in 80,000 Hours podcast he argued 'RL is evil' — JacquesThibs · 2026-09-24
- Is part of the population simply unable to recognize AI writing, even daily users? — robleclerc · 2026-09-24