SWE-agent

The 'SWE-agent' dataset/benchmark contains various multi-agent system scenarios used to evaluate the effectiveness of inter-agent communication strategies, particularly focusing on action-centered information preservation and performance-cost trade-offs.

Papers using SWE-agent (2)

What Should Agents Say? Action-state Communication for Efficient Multi-Agent Systems2026Coherence Collapse: Diagnosing Why Code Agents Fail After Reaching the Right Code2026