Isolated PR evaluations overestimate coding agent success rates by up to 20 percentage points, while agents that succeed still produce code with higher cognitive complexity and technical debt than human developers.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Beyond Isolated Tasks: A Framework for Evaluating Coding Agents on Sequential Software Evolution
Isolated PR evaluations overestimate coding agent success rates by up to 20 percentage points, while agents that succeed still produce code with higher cognitive complexity and technical debt than human developers.