---
title: "#mlsystems articles | Hashnode"
description: "Explore 2 articles tagged #mlsystems on Hashnode. Tutorials, guides, and how-tos from the community."
url: "https://hashnode.com/tag/mlsystems"
---

Tag

# #mlsystems

2 posts · 0 followers

[Daniel Anthony Romitelli Jr](https://hashnode.com/@romiteld) · [Aug 3](https://hashnode.com/posts/the-reward-calibrator-that-learns-the-shape-of-its-own-judgment/6a710461725ff6c213f38502)

### [The Reward Calibrator That Learns the Shape of Its Own Judgment](https://craftedbydaniel.hashnode.dev/the-reward-calibrator-that-learns-the-shape-of-its-own-judgment?utm_source=hashnode&utm_medium=feed)

I hit the wall when one noisy signal kept dragging candidate selection around like a shopping cart with one bad wheel. On paper, the composite score looked stable. Then I watched different weight sets

9 min read · [craftedbydaniel.hashnode.dev](https://craftedbydaniel.hashnode.dev/the-reward-calibrator-that-learns-the-shape-of-its-own-judgment?utm_source=hashnode&utm_medium=feed)

[Daniel Anthony Romitelli Jr](https://hashnode.com/@romiteld) · [Mar 10](https://hashnode.com/posts/phase-2-calibration-percategory-ood-thresholds-grouprelative-reward-normalization-in-my-scene-compiler/69b06c46b0b073f14fe9fa5f)

### [Phase 2 Calibration: Per‑Category OOD Thresholds + Group‑Relative Reward Normalization in My Scene Compiler](https://craftedbydaniel.hashnode.dev/phase-2-calibration-percategory-ood-thresholds-grouprelative-reward-normalization-in-my-scene-compiler?utm_source=hashnode&utm_medium=feed)

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...

12 min read · [craftedbydaniel.hashnode.dev](https://craftedbydaniel.hashnode.dev/phase-2-calibration-percategory-ood-thresholds-grouprelative-reward-normalization-in-my-scene-compiler?utm_source=hashnode&utm_medium=feed)

[Daniel Anthony Romitelli Jr](https://hashnode.com/@romiteld) · [Mar 10](https://hashnode.com/posts/mrgrpo-in-practice-the-reward-mixer-that-stops-clip-from-lying-to-your-scene-compiler/69b06bbb9616f3eaa481d77f)

### [MR‑GRPO in Practice: The Reward Mixer That Stops CLIP From Lying to Your Scene Compiler](https://craftedbydaniel.hashnode.dev/mrgrpo-in-practice-the-reward-mixer-that-stops-clip-from-lying-to-your-scene-compiler?utm_source=hashnode&utm_medium=feed)

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...

9 min read · [craftedbydaniel.hashnode.dev](https://craftedbydaniel.hashnode.dev/mrgrpo-in-practice-the-reward-mixer-that-stops-clip-from-lying-to-your-scene-compiler?utm_source=hashnode&utm_medium=feed)