Lay Paraphrasing on Five scientific datasets Mean (test)
3.82ComprehensivenessSci-LoRA
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Sci-LoRAFine-tuning=Dynamic Multi-LoRA Fusion2025.05 | 3.82 | 2.88 | 3.45 | 3.47 | 3.4 | 0.8 | |
| Qwen2.5Fine-tuning=LoRA (Fusion)2025.05 | 3.78 | 2.86 | 2.8 | 2.77 | 3.31 | 0.75 | |
| OPTFine-tuning=LoRA (Fusion)2025.05 | 3.65 | 2.13 | 2.69 | 2.33 | 2.46 | 0.76 | |
| Phi-4Fine-tuning=LoRA (Fusion)2025.05 | 3.53 | 2.77 | 2.27 | 2.2 | 2.71 | 0.75 | |
| MistralFine-tuning=LoRA (Fusion)2025.05 | 3.48 | 2.64 | 2.73 | 2.93 | 2.8 | 0.28 | |
| LLaMA3Fine-tuning=LoRA (Fusion)2025.05 | 3.47 | 3.1 | 2.73 | 2.72 | 2.87 | 0.72 | |
| MixtralMode=Pre-trained2025.05 | 3.42 | 2.67 | 2.72 | 2.89 | 2.64 | 0.64 | |
| GPT-4oMode=Pre-trained2025.05 | 3.25 | 2.68 | 3.08 | 3.53 | 3.15 | 0.75 |