Large language models develop novel social biases through adaptive exploration

hackernews 2026-09-09 원문 보기 ↗