Generative Classification on BigBench (test)
76.6AccuracyPrefixMemory-Tuning
Evaluation results
| Method | Links | |
|---|---|---|
| PrefixMemory-TuningBackbone Model=Qwen2.5-3B-Instruct2025.06 | 76.6 | |
| LoRA+Backbone Model=Qwen2.5-3B-Instruct2025.06 | 74.2 | |
| PrefixMemory-TuningBackbone Model=LLaMA2-7B-Chat2025.06 | 71.2 | |
| LoRA+Backbone Model=LLaMA2-7B-Chat2025.06 | 67.8 | |
| LoRABackbone Model=LLaMA2-7B-Chat, rank=r=642025.06 | 67.4 | |
| Full fine-tuningBackbone Model=Qwen2.5-3B-Instruct2025.06 | 67.4 | |
| LoRABackbone Model=Qwen2.5-3B-Instruct, rank=r=642025.06 | 61.4 | |
| Prefix-TuningBackbone Model=Qwen2.5-3B-Instruct, virtual tokens=m=322025.06 | 52 | |
| Full fine-tuningBackbone Model=LLaMA2-7B-Chat2025.06 | 38.8 | |
| Prefix-TuningBackbone Model=LLaMA2-7B-Chat, virtual tokens=m=322025.06 | 21.3 |