Dense Model Loss Curve Prediction WSD to Cosine transfer
0.202MAPECMMT (λ = 0.99)
Evaluation Results
| Method | Links | |
|---|---|---|
| CMMT (λ = 0.99)Model Size=100M2025.12 | 0.202 | |
| ASMTModel Size=100M2025.12 | 0.232 | |
| CMMT (λ = 0.99)Model Size=500M2025.12 | 0.379 | |
| CMMT (λ = 0.999)Model Size=500M2025.12 | 0.417 | |
| ASMTModel Size=500M2025.12 | 0.453 | |
| ASMTModel Size=50M2025.12 | 0.456 | |
| CMMT (λ = 0.99)Model Size=50M2025.12 | 0.539 | |
| CMMT (λ = 0.999)Model Size=50M2025.12 | 1.886 | |
| CMMT (λ = 0.999)Model Size=100M2025.12 | 2.251 |