List your company

Agentic Coding on MSWE-Bench

31.3ScoreSAR
25.26826.83428.429.966Jul 3, 2026

Evaluation results

MethodLinks
SARAgent=ClaudeCode, Projection=top-1% spectral-rank2026.0731.322.86
Baseline (In-house)Agent=CodeAct2026.0727.1—
SARAgent=CodeAct, Projection=top-1% spectral-rank2026.0727-0.33
Baseline (In-house)Agent=ClaudeCode2026.0725.5—