Two self-improving agent papers hit EMNLP: a 9B model's harness updates match Opus
yuyinzhou_cs · x · 2026-08-22
HenryLAI's team got two papers on self-improving agents accepted to EMNLP 2026, with code released in A-Evolve.
- "Harness Updating Is Not Harness Benefit": counter-intuitive findings — the updater's base capability barely matters; a Qwen3.5 9B model produces harness updates matching Claude Opus 4.6 (best-vs-worst evolver gap ≤3.1pp). Benefit is non-monotonic: mid-tier agents gain the most, strong models hit a ceiling, and weak models gain least (skill activation 25% vs 96% for strong) despite having the most headroom.
- "Evo-Harness: Context-to-Harness Skill Compilation": how to turn single noisy executions into reusable skill harnesses on the fly.
Both works tackle one core question: what actually makes an agent improve itself through its harness.
More from coding & agent
- Before building high-stakes AI systems: tell the LLM not to build anything — aryanXmahajan · 2026-08-22
- AI Refactoring Risks: Accidentally Deleting Valuable Technical Debt — sebpaquet · 2026-08-22
- MiniMax M3 ran overnight and found a TypeScript compiler crash—fuzzing first, then static analysis — DanielLockyer · 2026-08-22
- 10 Claude Code Steering Mechanisms That Stop Agents From Ignoring Instructions — bibryam · 2026-08-22
- DeepMind's Aletheia Agent Solves Math Conjectures via Self-Verification — thisguyknowsai · 2026-08-22
- Designing the Backend for Agent Systems: API, Queues, and Cost Control — kmeanskaran · 2026-08-22