SWE-INTERACT: Reimagining SWE Benchmarks as User-Driven Long-Horizon Coding Sessions
We introduce SWE-Interact, a new testbed for evaluating coding agents on multi-turn, interactive, user-driven software engineering tasks. Existing frontier SWE benchmarks typically provide complete requirements upfront and evaluate agents on autonomous implementation. In contrast, SWE-Interact places agents in a realistic developer workflow: a carefully designed user simulator starts with vague or incomplete instructions, progressively reveals requirements, inspects the agent's workspace, and provides targeted feedback, revisions, and new constraints until the full task goal has been handed of
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- LinkedLinked via unknownAgentTrace →
- LinkedLinked via unknownAgentCore-8B →
- LinkedLinked via unknownOpen-source agent framework crosses 50k stars →
- LinkedLinked via unknownagent-tools →
- LinkedLinked via unknownCursor now has a mobile app for guiding your coding agent on the go →
- PossiblePossibly related (embedding) · 29%SWE-agent/SWE-agent →
“Possibly related via embedding similarity 0.58 (not asserted). Timestamp check: artifact after paper (+16d).”
- LinkedLinked via unknownScarfBench: Benchmarking AI Agents for Enterprise Java Framework Migration →
- LinkedLinked via unknownREAP: Automatic Curation of Coding Agent Benchmarks from Interactive Production Usage [R] →