Prover-Verifier Games improve legibility of LLM outputs

Kirchner, Jan Hendrik, Chen, Yining, Edwards, Harri · arXiv · 2024