Software Engineering Repair on SWE Multi
40SWE Average ScoreRaw Mixture
Evaluation Results
| Method | Links | |
|---|---|---|
| Raw MixtureMid-training Protocol=Full Corpus, Token Budget=50B2026.05 | 40 | |
| MIRA-GroupMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 36.33 | |
| RandomMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 35 | |
| DataManMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 33 | |
| MIRA-GlobalMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 32 | |
| MIRA-SourceMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 30.33 | |
| DSIRMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 27 | |
| PPLMid-training Protocol=Subset Selection, Token Budget=25B2026.05 | 20 | |
| +SFTMid-training Protocol=No Mid-training, Token Budget=02026.05 | 3.67 | |
| Base ModelMid-training Protocol=No Mid-training, Token Budget=02026.05 | 0.33 |