Continual Learning on TRACE (test)
57.8Overall Performance ScoreSAE-FD
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| SAE-FDModel=Mistral-7B2026.05 | 57.8 | — | — | — | — | — | — | — | — | 1.63 | |
| MH-MoEBase model=Qwen3-8B2026.02 | 56.9 | 50.5 | 65.9 | 18.2 | 50.9 | 90.8 | 67.9 | 70.5 | 40.7 | -5.1 | |
| O-LoRABase model=Qwen3-8B2026.02 | 56.7 | 51.5 | 64.5 | 21.5 | 54.6 | 90.7 | 66.7 | 63.4 | 40.6 | -5.4 | |
| LoRAMoEBase model=Qwen3-8B2026.02 | 55.1 | 49.1 | 65.7 | 14.6 | 52.8 | 89.2 | 71.6 | 58.8 | 39.2 | -5.5 | |
| TreeLoRAModel=Mistral-7B2026.05 | 54.77 | — | — | — | — | — | — | — | — | 3.77 | |
| EWCBase model=Qwen3-8B2026.02 | 54.5 | 50.1 | 61.3 | 17.1 | 52.2 | 86.5 | 64.2 | 64.3 | 40.6 | -7.5 | |
| GEMBase model=Qwen3-8B2026.02 | 54.3 | 52.3 | 62.9 | 18 | 52.5 | 86.2 | 59.3 | 61.8 | 40.5 | -6.4 | |
| SeqLoRABase model=Qwen3-8B2026.02 | 54 | 49.6 | 56.7 | 16.7 | 49.7 | 88.8 | 66.7 | 63.7 | 40.2 | -7.3 | |
| SAE-FDModel=Vicuna-7B2026.05 | 53.5 | — | — | — | — | — | — | — | — | 0.69 | |
| SAE-FDModel=LLaMA-2-7B2026.05 | 52.7 | — | — | — | — | — | — | — | — | 0.46 | |
| EWCModel=Mistral-7B2026.05 | 52.45 | — | — | — | — | — | — | — | — | 5.98 | |
| O-LoRAModel=Mistral-7B2026.05 | 52.02 | — | — | — | — | — | — | — | — | 8.13 | |
| GORPModel=LLaMA-2-7B2026.05 | 50.4 | — | — | — | — | — | — | — | — | 0.7 | |
| MH-MoEBase model=Qwen3-0.6B2026.02 | 46.7 | 48.8 | 67.3 | 18.8 | 49.3 | 70.1 | 34.6 | 48.3 | 36.7 | -4.5 | |
| TreeLoRAModel=LLaMA-2-7B2026.05 | 43.52 | — | — | — | — | — | — | — | — | 3.46 | |
| O-LoRAModel=Vicuna-7B2026.05 | 43.42 | — | — | — | — | — | — | — | — | 6.27 | |
| O-LoRABase model=Qwen3-0.6B2026.02 | 43.4 | 46.3 | 54.4 | 20.9 | 47.9 | 65.3 | 27.2 | 48.3 | 37.2 | -8.1 | |
| O-LoRAModel=LLaMA-2-7B2026.05 | 42.78 | — | — | — | — | — | — | — | — | 7.16 | |
| EWCModel=LLaMA-2-7B2026.05 | 42.36 | — | — | — | — | — | — | — | — | 5.97 | |
| EWCModel=Vicuna-7B2026.05 | 41.88 | — | — | — | — | — | — | — | — | 15.57 | |
| LwFModel=Vicuna-7B2026.05 | 41.19 | — | — | — | — | — | — | — | — | 5.54 | |
| LoRAMoEBase model=Qwen3-0.6B2026.02 | 37.8 | 34.3 | 52 | 12 | 46.3 | 61.4 | 14.8 | 44.3 | 37.2 | -11.2 | |
| GEMBase model=Qwen3-0.6B2026.02 | 35.7 | 46.1 | 30.8 | 17.7 | 45.5 | 53.2 | 19.8 | 35.7 | 36.9 | -12.4 | |
| EWCBase model=Qwen3-0.6B2026.02 | 35.5 | 46.6 | 42.3 | 17.6 | 45.8 | 48.2 | 8.6 | 37.5 | 37 | -12.3 | |
| SeqLoRABase model=Qwen3-0.6B2026.02 | 29.5 | 12.8 | 26.2 | 14.6 | 43.5 | 59.8 | 7.4 | 34.8 | 36.6 | -19 |