Continual Learning on Long CL benchmark N=15
76.7Long1 PerformancePS-LoRA
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| PS-LoRABackbone=meta-llama/Llama-2-7b-chat2025.11 | 76.7 | 76.1 | 76.2 | 0.763 | |
| MoCLBackbone=meta-llama/Llama-2-7b-chat2025.11 | 75.2 | 70.7 | 74.8 | 0.736 | |
| PS-LoRABackbone=google-t5/t5-large2025.11 | 74.2 | 76.5 | 75.7 | 0.755 | |
| AM-LoRABackbone=google-t5/t5-large2025.11 | 72.7 | 73.3 | 71.8 | 0.726 | |
| O-LoRABackbone=google-t5/t5-large2025.11 | 71.5 | 66.7 | 71.3 | 0.698 | |
| MIGUBackbone=google-t5/t5-large2025.11 | 71.3 | 67.7 | 67.3 | 0.687 | |
| MIGUBackbone=meta-llama/Llama-2-7b-chat2025.11 | 71.2 | 70.6 | 70.5 | 0.705 | |
| OLoRABackbone=meta-llama/Llama-2-7b-chat2025.11 | 71.1 | 68.9 | 73.8 | 0.713 | |
| IncLoRABackbone=meta-llama/Llama-2-7b-chat2025.11 | 70.7 | 70.8 | 69.2 | 0.702 | |
| SD-LoRABackbone=meta-llama/Llama-2-7b-chat2025.11 | 70.2 | 68.4 | 70.9 | 0.698 | |
| MoCLBackbone=google-t5/t5-large2025.11 | 69.6 | 70.2 | 70.9 | 0.702 | |
| SD-LoRABackbone=google-t5/t5-large2025.11 | 69.3 | 69.8 | 70 | 0.697 | |
| SeqLoRABackbone=meta-llama/Llama-2-7b-chat2025.11 | 69 | 70.5 | 66.9 | 0.688 | |
| IncLoRABackbone=google-t5/t5-large2025.11 | 60.3 | 60.5 | 53.2 | 0.58 | |
| L2PBackbone=google-t5/t5-large2025.11 | 57.5 | 53.8 | 56.9 | 0.561 | |
| LwFBackbone=google-t5/t5-large2025.11 | 50.1 | 43.1 | 47.4 | 0.469 | |
| EWCBackbone=google-t5/t5-large2025.11 | 45.3 | 44.5 | 45.6 | 0.451 | |
| SeqLoRABackbone=google-t5/t5-large2025.11 | 12.3 | 10.1 | 10.1 | 0.108 |