Harnesses Optimization
Properties
created
14.07.2026, 23:04
modified
14.07.2026, 23:06
published
Empty
sources
Empty
topics
Agentic Harness, Harness Optimization
authors
Jakub
ai-assisted
No
- This note aggregates information about optimizing harnesses
# Papers
- Meta-Harness
- 2026 — Stanford paper introducing the outer-loop harness-search framework: a coding-agent proposer reads prior code, scores, and traces from a filesystem to propose new harnesses; beats hand-designed ACE by 7.7 points on text classification and ranks #1 on TerminalBench-2 Haiku 4.5
- JoelNiklaus/harness-optimization
- Self-Harness
- 2026-06-14 — agent improves its own harness (prompts, tools, memory) via weakness mining → proposal → validation
- Autogenesis
- 2026-04-26 — two-layer self-evolving agent protocol with auditable lineage and rollback
- Red Queen Gödel Machine
- 2026-07-05 — co-evolves agents and the evaluators that judge them to escape the stationary-evaluator plateau
- Don’t Train the Model, Evolve the Harness
- 2026-07-01 — Hugging Face applies Meta-Harness to DeepSeek-V4-Pro on Harvey’s Legal Agent Benchmark: 0% → 5.0% all-pass, 63.4% → 80.1% criterion pass on held-out test, zero weights changed