
Qwen3.8-27B, 600 steps, 12 reward channels: CISPO, REPO-R and prompt filtering lifted holdout from −0.47 to +1.80, and an advantage floor silently zeroed 96% of the signal.
Ex-Apple GenAI engineer, building an LLM post-training platform at g factor technologies
Ex-Apple GenAI engineer, building an LLM post-training platform at g factor technologies
Markdown supported · ⌘↵ to post
No responses yet.