Coding Agents Don't Know When to Act
For developers deploying autonomous coding agents, this work highlights a critical failure mode where agents introduce technical debt by modifying already-fixed code, and shows that current models lack the ability to appropriately abstain from action.
Coding agents often fail to recognize when a reported bug has already been fixed, proposing unnecessary code changes in 35-65% of cases. The study introduces FixedBench, a benchmark of 200 tasks requiring no changes, and finds that even state-of-the-art models exhibit an action bias, preferring to act even when inaction is appropriate.
Coding agents are increasingly deployed to autonomously maintain software, including to resolve user-reported issues: a bug report comes in and the agent creates a patch to address it. However, in any real-world deployment, they will encounter stale bug reports about issues that have already been resolved. Agents should recognize this and abstain from modifying the code to avoid accumulating technical debt. To systematically evaluate whether current agents do so, we introduce FixedBench, a code benchmark with 200 human-verified coding tasks in which no code changes are required, testing five recent models across four agent harnesses. We find that even state-of-the-art models fail, proposing undesirable changes (excluding tests and documentation) in $35$ to $65\%$ of cases. Explicit instructions to reproduce the issue before patching partially address this issue but introduce a new failure mode: when an issue is partially fixed, they abstain even though a patch would still be needed. More broadly, our results indicate that LLMs fall prey to an action bias: they choose to act even if inaction would be appropriate. To break this pattern, inaction needs to be explicitly framed as a path to success, which highlights an overreliance on human guidance implicit in current training objectives.