Classification on MultiRC
82.6AccuracyFirst Order Adamw FT
Evaluation results
| Method | Links | |
|---|---|---|
| First Order Adamw FTModel Backbone=Llama2-7B, Fine-tuning Strategy=Full Fine-tuning, Optimization Algorithm=First Order Adamw, Number of training examples=10002026.06 | 82.6 | |
| FTOptimization Algorithm=Adam, Tuning Strategy=Full Tuning, Model=OPT-13B2026.01 | 71.1 | |
| Dominant-layer ZO LoRAModel Backbone=Llama2-7B, Fine-tuning Strategy=LoRA, Optimization Algorithm=Dominant-layer ZO, Number of training examples=10002026.06 | 66.52 | |
| MeZO LoRAModel Backbone=Llama2-7B, Fine-tuning Strategy=LoRA, Optimization Algorithm=MeZO, Number of training examples=10002026.06 | 65.9 | |
| Dominant-layer ZO FTModel Backbone=Llama2-7B, Fine-tuning Strategy=Full Fine-tuning, Optimization Algorithm=Dominant-layer ZO, Number of training examples=10002026.06 | 65.86 | |
| SubZero-GV (Prefix)Optimization Algorithm=SubZero, Tuning Strategy=Prefix Tuning, Guiding Vectors=true, Model=OPT-13B2026.01 | 65.1 | |
| HiZOO (Prefix)Optimization Algorithm=HiZOO, Tuning Strategy=Prefix Tuning, Model=OPT-13B2026.01 | 64.8 | |
| MeZO-GV (Prefix)Optimization Algorithm=MeZO, Tuning Strategy=Prefix Tuning, Guiding Vectors=true, Model=OPT-13B2026.01 | 64.4 | |
| MeZO FTModel Backbone=Llama2-7B, Fine-tuning Strategy=Full Fine-tuning, Optimization Algorithm=MeZO, Number of training examples=10002026.06 | 64.4 | |
| SubZero-GV (LoRA)Optimization Algorithm=SubZero, Tuning Strategy=LoRA, Guiding Vectors=true, Model=OPT-13B2026.01 | 64.1 | |
| MeZO (Prefix)Optimization Algorithm=MeZO, Tuning Strategy=Prefix Tuning, Guiding Vectors=false, Model=OPT-13B2026.01 | 63.7 | |
| SubZero (Prefix)Optimization Algorithm=SubZero, Tuning Strategy=Prefix Tuning, Guiding Vectors=false, Model=OPT-13B2026.01 | 63.5 | |
| SubZero-GV (FT)Optimization Algorithm=SubZero, Tuning Strategy=Full Tuning, Guiding Vectors=true, Model=OPT-13B2026.01 | 63.2 | |
| ZO-AdaMU (2x)Optimization Algorithm=ZO-AdaMU, Tuning Strategy=Full Tuning, Training Steps=2x, Model=OPT-13B2026.01 | 63 | |
| MeZO-GV(Prefix)Backbone=OPT-1.3B, Optimization Method=MeZO, Adaptation Technique=Prefix Tuning, Guiding Vectors (GV)=true2026.01 | 62.7 | |
| MeZO-GV (FT)Optimization Algorithm=MeZO, Tuning Strategy=Full Tuning, Guiding Vectors=true, Model=OPT-13B2026.01 | 62.5 | |
| ZO-AdaMU (Prefix)Optimization Algorithm=ZO-AdaMU, Tuning Strategy=Prefix Tuning, Model=OPT-13B2026.01 | 61.9 | |
| MeZO-GV (LoRA)Optimization Algorithm=MeZO, Tuning Strategy=LoRA, Guiding Vectors=true, Model=OPT-13B2026.01 | 61.9 | |
| MeZO (LoRA)Optimization Algorithm=MeZO, Tuning Strategy=LoRA, Guiding Vectors=false, Model=OPT-13B2026.01 | 61.5 | |
| HiZOOOptimization Algorithm=HiZOO, Tuning Strategy=Full Tuning, Model=OPT-13B2026.01 | 61.3 | |
| SubZero (FT)Optimization Algorithm=SubZero, Tuning Strategy=Full Tuning, Guiding Vectors=false, Model=OPT-13B2026.01 | 61 | |
| MeZO-GV(LoRA)Backbone=OPT-1.3B, Optimization Method=MeZO, Adaptation Technique=LoRA, Guiding Vectors (GV)=true2026.01 | 60.6 | |
| SubZero (LoRA)Optimization Algorithm=SubZero, Tuning Strategy=LoRA, Guiding Vectors=false, Model=OPT-13B2026.01 | 60.3 | |
| HiZOO (LoRA)Optimization Algorithm=HiZOO, Tuning Strategy=LoRA, Model=OPT-13B2026.01 | 60.2 | |
| MeZO (FT)Optimization Algorithm=MeZO, Tuning Strategy=Full Tuning, Guiding Vectors=false, Model=OPT-13B2026.01 | 60.1 | |
| MeZO(Prefix)Backbone=OPT-1.3B, Optimization Method=MeZO, Adaptation Technique=Prefix Tuning2026.01 | 59.1 | |
| ZO-AdaMU (LoRA)Optimization Algorithm=ZO-AdaMU, Tuning Strategy=LoRA, Model=OPT-13B2026.01 | 58.9 | |
| MeZO-GV(FT)Backbone=OPT-1.3B, Optimization Method=MeZO, Adaptation Technique=Fine-tuning, Guiding Vectors (GV)=true2026.01 | 58.6 | |
| MeZO(LoRA)Backbone=OPT-1.3B, Optimization Method=MeZO, Adaptation Technique=LoRA2026.01 | 57 | |
| MeZO(FT)Backbone=OPT-1.3B, Optimization Method=MeZO, Adaptation Technique=Fine-tuning2026.01 | 53.3 | |
| ICLEvaluation Protocol=In-context Learning, Model=OPT-13B2026.01 | 53.1 | |
| Zero-shotEvaluation Protocol=Zero-shot, Model=OPT-13B2026.01 | 46.9 | |
| ICLBackbone=OPT-1.3B, Learning Strategy=In-context Learning2026.01 | 46.3 | |
| Zero-shotBackbone=OPT-1.3B, Learning Strategy=Zero-shot2026.01 | 45.4 | |
| Zero-shot w/o finetuneModel Backbone=Llama2-7B, Fine-tuning Strategy=None, Optimization Algorithm=None, Number of training examples=10002026.06 | 45.3 |