G github.com My LLM optimization loop reward-hacked its own benchmark (and other lessons) [pdf] 1 points CodeReclaimers 5 months ago 2 comments