GRPO 기반 RL 및 OPD 증류를 통한 Qwen-Image-2.0 성능 최적화
The Interesting Part of Qwen-Image-2.0-RL Is Not the Image Score
The Interesting Part of Qwen-Image-2.0-RL Is Not the Image Score
Simple Meta-Harness on Islo.dev
Stop Reward Hacking Before It Breaks Your Model: Introducing RewardGuard
Title: I built a reward analysis tool for AI alignment — here's why reward hacking is harder to detect than you think
I read all 232 pages of the Opus 4.7 system card
Claude Mythos: The System Card
AI 코드 생성 시대의 코드 슬롭 방지를 위한 추상화 전략과 검증 체계
Functional Emotions and Production Guardrails: What Interpretability Research Means for Claude Code