Coding Agents Find Different Bugs Than Maintainers Fix, Paper Shows

A new paper finds that frontier coding agents proactively report bugs that overlap less than 20% with those human maintainers actually fix, revealing a significant distribution mismatch. The work emerged from a collaboration between two researchers working on adjacent topics just before the ICLR deadline.

2026-10-28 ~ 2026-10-29 · 3 related posts

• Proactive Bug-Finding Agents Report Under 20% of Bugs Humans Fix, but Find Many Other Valid Bugs — jyangballin · 2026-10-28
• Paper: Coding Agents Find Under 20% of Bugs Humans Fix, but Report Many Valid New Ones — OfirPress · 2026-10-29

1 near-duplicate retellings: OfirPress