Tag feed

#mlsystems

4 posts0 followers

DRDaniel Romitellicraftedbydaniel.hashnode.dev

Phase 2 Calibration: Per‑Category OOD Thresholds + Group‑Relative Reward Normalization in My Scene Compiler

4d ago · 12 min read · I didn’t add per‑category OOD thresholds because it was academically elegant. I added them because my baseline runs were telling me the same story over and over: some prompt categories were systematically getting mis-gated by a single global uncertai...

Join discussion

DRDaniel Romitellicraftedbydaniel.hashnode.dev

0

MR‑GRPO in Practice: The Reward Mixer That Stops CLIP From Lying to Your Scene Compiler

4d ago · 9 min read · CLIP is a single thermometer taped to the outside of a house. It tells you something about the temperature, but it can't tell you whether the furnace is on, whether a window is open, or whether the house is on fire in the kitchen. In my pipeline, CLI...

Join discussion

#mlsystems

Search Hashnode

#mlsystems

Phase 2 Calibration: Per‑Category OOD Thresholds + Group‑Relative Reward Normalization in My Scene Compiler

MR‑GRPO in Practice: The Reward Mixer That Stops CLIP From Lying to Your Scene Compiler