File-level Localization on MuLocBench
29.2Top-1 AccuracyClaude Code*
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Claude Code*Category=Agent-based, Backbone=Claude-Opus-4.62026.05 | 29.2 | 54.9 | 41.4 | 69 | |
| CodexCategory=Agent-based, Backbone=GPT-5.22026.05 | 28 | 50 | 39.8 | 65.1 | |
| LARGERTunedCategory=Agent-based (graph-enhanced), Backbone=GPT-5.2, Setting=Tuned2026.05 | 28 | 60 | 40 | 72 | |
| LARGERFixedCategory=Agent-based (graph-enhanced), Backbone=GPT-5.2, Setting=Fixed2026.05 | 27.4 | 55.7 | 39.4 | 68.6 | |
| OpenHandsCategory=Agent-based, Backbone=GPT-5.22026.05 | 27.3 | 43.7 | 38 | 57.6 | |
| mini-swe-agentCategory=Agent-based, Backbone=GPT-5.22026.05 | 26.9 | 46.6 | 38.1 | 61.4 | |
| SWE-agentCategory=Agent-based, Backbone=GPT-5.22026.05 | 21 | 31.3 | 30.4 | 44.5 | |
| AgentlessCategory=Procedure-based, Backbone=GPT-5.22026.05 | 20.4 | 28.9 | 29.5 | 42 | |
| CoSILCategory=Agent-based (graph-enhanced), Backbone=GPT-5.22026.05 | 18.1 | 27 | 27 | 40.1 | |
| LocAgentCategory=Agent-based (graph-enhanced), Backbone=GPT-5.22026.05 | 17.9 | 25.7 | 27.3 | 39 | |
| BM25Category=Lexical retrieval2026.05 | 10.6 | 24.5 | 16.4 | 35.3 |