Test Writing on SWE-Atlas
37.78ScoreLARGERFixed
Evaluation Results
| Method | Links | |
|---|---|---|
| LARGERFixedBackbone LLM=GPT-5.22026.05 | 37.78 | |
| Claude Code*Backbone LLM=Claude Opus 4.62026.05 | 36.67 | |
| CodexBackbone LLM=GPT-5.22026.05 | 32.22 | |
| mini-swe-agentBackbone LLM=GPT-5.22026.05 | 27.78 |