SWE-chat is the first large-scale dataset of real-world AI coding agent sessions from open-source developers, revealing that only 44% of agent-written code survives into commits and that users push back in 44% of turns.
score": <integer 0-100>,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SWE-chat: Coding Agent Interactions From Real Users in the Wild
SWE-chat is the first large-scale dataset of real-world AI coding agent sessions from open-source developers, revealing that only 44% of agent-written code survives into commits and that users push back in 44% of turns.