Meta-Harness: End-to-End Optimization of Model Harnesses
Source
yoonholee.com
Author
Yoonho Lee
Date
Why it matters
Harness engineering, automated: this optimizes the scaffolding around a model — what it stores, retrieves, and feeds back — rather than the weights.
Read it once you have realized the harness, not the model, is where most of your agent quality actually comes from.
A research paper and framework demonstrating automated optimization of model harnesses - the scaffolding code that determines what AI models store, retrieve, and present during task execution.
Shows significant improvements over hand-designed systems across text classification, math reasoning, and coding tasks.
Transcript
We just released code for Meta-Harness! https://t.co/OdU7zocdPl
Aside from replicating paper experiments, the repo is designed to help users implement good Meta-Harnesses in completely new domains! Just point your agent at ONBOARDING.md and have a conversation https://t.co/0H6Zrvg8FQ https://t.co/0hhtPHT1Ut