Saddle scrambling lets reasoning models explore more solutions, measured by Lyapunov indicator
wgilpin0 · x · 2026-09-09
The team explains the mechanism: high-dimensional dynamics act like a Plinko game — trajectories start nearby and converge to the same answer, but saddles scatter them along different routes via incorrect answers. 'Saddle scrambling' lets reasoning models check more possible solutions. They compute the fast Lyapunov indicator to measure transient chaos, finding it correlates with how many solutions the model visits before converging.
Related event: Fractal basins and transient chaos explain why reasoning models overthink(13 posts)→
More from Research
- Terence Tao: The Question Actually Had a Small Finite Counterexample — tak3sh8 · 2026-09-09
- Uno speeds up Qwen3-8B 2.5x by using diffusion for parallel token drafting — rohanpaul_ai · 2026-09-09
- 4.9M messages, 300B tokens, 10,000 agents: details of OpenAI's Navier–Stokes run — willdepue · 2026-09-09
- OpenAI's Navier-Stokes breakthrough: 10,000 agents solved it in 88 hours — generativist · 2026-09-09
- 16,893 sessions measured: what Claude Code, Codex and Cursor actually recommend — rseroter · 2026-09-09
- Largest open experiment yet on autonomous agents iterating in a research environment — eliebakouch · 2026-09-09