Mathematical Reasoning on SVAMP (Exact-match Accuracy)
66.12Exact Match AccuracyFull-data Fine-tuning
Evaluation Results
| Method | Links | |
|---|---|---|
| Full-data Fine-tuningBackbone=LLAMA-2-7B, Coreset Method=Full-data, Coreset Fraction=100%2025.10 | 66.12 | |
| LESSBackbone=LLAMA-2-7B, Coreset Method=LESS, Coreset Fraction=50%2025.10 | 65.45 | |
| S2LBackbone=LLAMA-2-7B, Coreset Method=S2L, Coreset Fraction=50%2025.10 | 65.3 | |
| TRIMBackbone=LLAMA-2-7B, Coreset Method=TRIM, Coreset Fraction=50%2025.10 | 65.12 | |
| TAGCOSBackbone=LLAMA-2-7B, Coreset Method=TAGCOS, Coreset Fraction=50%2025.10 | 64.15 | |
| RandomBackbone=LLAMA-2-7B, Coreset Method=Random, Coreset Fraction=50%2025.10 | 63.3 | |
| Pretrained (no Fine-tuning)Backbone=LLAMA-2-7B, Coreset Method=Pretrained2025.10 | 12.4 |