Topic
mini-SWE-agent
Latest news
- 24 OctStudy finds fancy agent harnesses raise cost but not reliable performance gains
A study tested five models across three agent harnesses and found much of the reported performance gap comes from run-to-run noise and wiring differences rather than real improvements. The researchers also observed that web access allowed some models to copy fixes from GitHub, so they ran tests offline to remove that effect.
Research · 1 source

Articles
No articles about mini-SWE-agent yet.