Reading Comprehension on DROP (Loss and F1)
0.4LossZO Fine-tuner
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ZO Fine-tunerModel=Qwen2.5-14B2025.10 | 0.4 | 70 | |
| MeZOModel=Qwen2.5-14B2025.10 | 0.45 | 66 | |
| LOZOModel=Qwen2.5-14B2025.10 | 0.46 | 67 | |
| HIZOOModel=Qwen2.5-14B2025.10 | 0.49 | 68 | |
| MeZO-AdamUModel=Qwen2.5-14B2025.10 | 0.5 | 64 | |
| ZO Fine-tunerModel=LLaMA-3.1-8B2025.10 | 0.54 | 66 | |
| HIZOOModel=LLaMA-3.1-8B2025.10 | 0.64 | 60 | |
| MeZOModel=LLaMA-3.1-8B2025.10 | 0.69 | 64 | |
| MeZO-AdamUModel=LLaMA-3.1-8B2025.10 | 0.73 | 59 | |
| LOZOModel=LLaMA-3.1-8B2025.10 | 0.74 | 65 | |
| ZO Fine-tunerModel=LLaMA-3.2-1B2025.10 | 1.03 | 35 | |
| HIZOOModel=LLaMA-3.2-1B2025.10 | 1.09 | 29 | |
| MeZO-AdamModel=LLaMA-3.2-1B2025.10 | 1.14 | 29 | |
| LOZOModel=LLaMA-3.2-1B2025.10 | 1.15 | 32 | |
| MeZOModel=LLaMA-3.2-1B2025.10 | 1.16 | 29 | |
| ZO Fine-tunerModel=OPT-30B2025.10 | 1.59 | 31 | |
| HIZOOModel=OPT-30B2025.10 | 1.61 | 30 | |
| MeZOModel=OPT-30B2025.10 | 1.66 | 31 | |
| MeZO-AdamUModel=OPT-30B2025.10 | 1.7 | 30 | |
| LOZOModel=OPT-30B2025.10 | 2.83 | 27 |