SGLang SOTA Humanize Loop

Overview

Use this skill when the user names a model and wants the SGLang serving path to autonomously keep improving until it matches or beats the best reproducible result from the requested comparison framework set in the same target environment.

This workflow has two durable parts:

1. A fixed baseline phase that must be completed once before any code patching.
2. One Humanize RLCR loop that owns gap decision, profiling, required layer/kernel deep dive, SGLang patching, optional NCU evidence, and real-model revalidation.