Reasoning models 'overthink': similar prompts yield 10x difference in reasoning length and cost
wgilpin0 · x · 2026-09-09
In a thread, the authors show reasoning models commonly 'overthink': two similar prompts can produce reasoning traces differing by 10x in length, scaling token cost. They parametrize problem + weights as a dynamical system whose solution is a fixed point, and by varying initial latent states across a random 2D slice, they find fractal basins in convergence time.
More from Research
- LightMTP: lightweight latent multi-token prediction boosts LLM planning in pretraining — gdm3000 · 2026-10-27
- EMNLP study: AI literary translation reads fine but varies more, even within one book — tuvllms · 2026-10-27
- Lab monkeys at $30-80K each: a 4-year breeding cycle is the real preclinical drug bottleneck — zebird0 · 2026-10-27
- OpenAI Foundation funds UNC tumor antigen project, hiring senior experimental scientist — iskander · 2026-10-27
- Steering attention heads toward honesty: up moral patienthood, down psychopathy — edelwax · 2026-10-27
- Are transcoders overkill? New paper argues plain neurons suffice for circuit analysis — burny_tech · 2026-10-27