Quality review for benchmarks and post-training data

We thoroughly review your benchmarks, post-training data, reinforcement learning environments and more to find flaws and recommend improvements.