Multi-domain Evaluation on MMLU, GSM8K, HEval, BBH, MedQA
68Average ScoreCODA
Evaluation Results
| Method | Links | |
|---|---|---|
| CODABackbone=Qwen-2.5-7B, Data Budget=50%2026.07 | 68 | |
| GradNormBackbone=Qwen-2.5-7B, Data Budget=50%2026.07 | 66.8 | |
| GREATSBackbone=Qwen-2.5-7B, Data Budget=50%2026.07 | 66.7 | |
| PCGradBackbone=Qwen-2.5-7B, Data Budget=50%2026.07 | 66.6 | |
| DoReMiBackbone=Qwen-2.5-7B, Data Budget=50%2026.07 | 66.4 | |
| Full dataBackbone=Qwen-2.5-7B, Data Budget=100%2026.07 | 66.2 | |
| TemperatureBackbone=Qwen-2.5-7B, Data Budget=50%2026.07 | 66 | |
| ProportionalBackbone=Qwen-2.5-7B, Data Budget=50%2026.07 | 65.7 | |
| Uniform mixingBackbone=Qwen-2.5-7B, Data Budget=50%2026.07 | 65.5 | |
| CODABackbone=LLaMA-3.1-8B, Data Budget=50%2026.07 | 57.2 | |
| GradNormBackbone=LLaMA-3.1-8B, Data Budget=50%2026.07 | 56 | |
| GREATSBackbone=LLaMA-3.1-8B, Data Budget=50%2026.07 | 55.8 | |
| PCGradBackbone=LLaMA-3.1-8B, Data Budget=50%2026.07 | 55.7 | |
| DoReMiBackbone=LLaMA-3.1-8B, Data Budget=50%2026.07 | 55.5 | |
| Full dataBackbone=LLaMA-3.1-8B, Data Budget=100%2026.07 | 55.4 | |
| TemperatureBackbone=LLaMA-3.1-8B, Data Budget=50%2026.07 | 55.1 | |
| ProportionalBackbone=LLaMA-3.1-8B, Data Budget=50%2026.07 | 54.7 | |
| Uniform mixingBackbone=LLaMA-3.1-8B, Data Budget=50%2026.07 | 54.5 |