Language Understanding on MMLU (T1)
48.1AccuracyRandom Rehearsal
Evaluation Results
| Method | Links | |
|---|---|---|
| Random RehearsalBackbone=Phi-4-Mini-Instruct-3.8B2026.04 | 48.1 | |
| Naive Fine-tuningBackbone=Phi-4-Mini-Instruct-3.8B2026.04 | 47.9 | |
| EWCBackbone=Phi-4-Mini-Instruct-3.8B2026.04 | 47.7 | |
| COMPASS-ECDABackbone=Phi-4-Mini-Instruct-3.8B2026.04 | 47.7 | |
| Full RetrainingBackbone=Phi-4-Mini-Instruct-3.8B2026.04 | 47.5 | |
| Naive Fine-tuningModel=Qwen2.5-7B-Instruct2026.04 | 0.604 | |
| EWCModel=Qwen2.5-7B-Instruct2026.04 | 0.603 | |
| Full RetrainingModel=Qwen2.5-7B-Instruct2026.04 | 0.6 | |
| Random RehearsalModel=Qwen2.5-7B-Instruct2026.04 | 0.6 | |
| COMPASS-ECDAModel=Qwen2.5-7B-Instruct2026.04 | 0.6 |