This repository is the public sample and schema repository for Ulam's math olympiad reasoning trajectories for reinforcement learning with verifiable rewards (RLVR), answer-verifier evaluation, process-supervision candidates, judge training, proof criticism, and private evaluations.
The goal is not merely to provide final-answer math examples. Each record is a structured olympiad reasoning object containing a normalized problem… See the full description on the dataset page:
https://huggingface.co/datasets/ulamai/verified-math-olympiad-trajectories.
