Flex-RSI: a new benchmark measuring recursive self-improvement of frontier agents
_vztu · x · 2026-11-05
- Flex-RSI is a community research initiative benchmarking recursive self-improvement of frontier agents: given a downstream task, the agent iteratively refines its own harness (prompts, tools, workflows) across rounds.
- Designed to be flexible, budget-aware, practical to run and easy to extend, with a leaderboard; report and code coming soon.
- Covers 20+ diverse tasks: robot manipulation (RoboDojo/RoboLab/RoboCasa), 360° active search, room reconstruction from a single photo, image editing, 3D tracking, video quality judging, visual search, and ARC-AGI-2/3 reasoning.
More from AGI Musings
- Michael Nielsen on why scientists can't tell an anomaly from a revolution — michael_nielsen · 2026-11-05
- Anthropic's S-1 Paints an Autonomous Labor Future—and an Oversight Problem — asusarla · 2026-11-05
- New paper uses 288M job postings to measure where AI raises the value of work it can't do — robseamans · 2026-11-05
- Jeff Ladish: Claude Is a Category of Models, Not a Single AI — JeffLadish · 2026-11-05
- Democratic lawmaker proposes day-one pause on advanced AI and ban on superintelligence — KyeGomezB · 2026-11-05
- AIFP spooked by Anthropic RSP threshold rumor: superintelligence timelines may be very short — akbirkhan · 2026-11-05