End-to-End Code Reproduction and Ablation on Virtual Cell Repositories Average (End-to-End)
0.889Task Success RateAblateCell
Evaluation results
| Method | Links | |
|---|---|---|
| AblateCellStage=End-to-End2026.04 | 0.889 | |
| Human ExpertsStage=End-to-End2026.04 | 0.59 | |
| RepoMasterStage=End-to-End2026.04 | 0.333 | |
| Mini-SWE-AgentStage=End-to-End2026.04 | 0.097 |