open-cot

open-cot provides Python research scripts for reproducing published methods that elicit reasoning-like text from language models through API calls. It includes individual sessions, benchmark batches and reasoning-effort comparisons, with reports against problems having programmatically checked answers. Researchers run the scripts with their own API configuration; the documented extracted text is treated as a proxy for internal reasoning.

open-cot: official README · Reviewed 2026-10-22