Medical information extraction and understanding on MedMCQA
3.28Perplexity (PPL)Base
Evaluation results
| Method | Links | ||
|---|---|---|---|
| BaseBackbone=GPT-J2026.04 | 3.28 | 21.3 | |
| RSLoRABackbone=GPT-J2026.04 | 3.31 | 35.69 | |
| Selective Parameter OptimizationBackbone=GPT-J2026.04 | 3.36 | 36.18 | |
| EWCLoRABackbone=GPT-J2026.04 | 3.38 | 36.48 | |
| LoRA(ν + μ)Backbone=GPT-J2026.04 | 3.41 | 37.13 | |
| LoRA(μ)Backbone=GPT-J2026.04 | 3.49 | 38.35 | |
| Selective Parameter OptimizationBackbone=LLaMA-32026.04 | 4.62 | 55.88 | |
| EWCLoRABackbone=LLaMA-32026.04 | 4.79 | 56.39 | |
| BaseBackbone=LLaMA-32026.04 | 4.94 | 18.5 | |
| LoRA(μ)Backbone=LLaMA-32026.04 | 5.04 | 53.69 | |
| LoRA(ν + μ)Backbone=LLaMA-32026.04 | 5.15 | 53.12 | |
| RSLoRABackbone=LLaMA-32026.04 | 5.24 | 51.92 |