Text Classification on SUBJ
93.36CA (%)ICL
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ICLBackbone=Qwen-3-8B2026.05 | 93.36 | — | — | — | |
| LTVBackbone=Qwen-3-8B2026.05 | 90 | — | — | — | |
| ICLBackbone=Qwen-2.5-7B2026.05 | 89.68 | — | — | — | |
| ICLBackbone=LLaMA-3.1-8B2026.05 | 85.96 | — | — | — | |
| ICLBackbone=LLaMA-2-13B2026.05 | 82.28 | — | — | — | |
| TATRADataset-free=true2026.02 | 82.18 | — | — | — | |
| PIAST (E)Dataset-free=false2026.02 | 80.98 | — | — | — | |
| LTVBackbone=Qwen-2.5-7B2026.05 | 80.68 | — | — | — | |
| LTVBackbone=LLaMA-2-13B2026.05 | 78.76 | — | — | — | |
| BCModel=Qwen, k-shot=8, Seeds=52025.05 | 77.03 | — | — | — | |
| PRLDataset-free=false2026.02 | 76.9 | — | — | — | |
| PRLBackbone=Qwen2.5-7B-Instruct, Prompt Selection=True2025.05 | 76.9 | — | — | — | |
| PromptAgentBackbone=Qwen2.5-7B-Instruct2025.05 | 76.83 | — | — | — | |
| Task VectorBackbone=Qwen-3-8B2026.05 | 76 | — | — | — | |
| PIASTDataset-free=false2026.02 | 75.93 | — | — | — | |
| State VectorBackbone=Qwen-3-8B2026.05 | 75.56 | — | — | — | |
| GADataset-free=false2026.02 | 74.93 | — | — | — | |
| GABackbone=Qwen2.5-7B-Instruct2025.05 | 74.93 | — | — | — | |
| APEDataset-free=false2026.02 | 73.92 | — | — | — | |
| GRACEDataset-free=false2026.02 | 73.92 | — | — | — | |
| APEBackbone=Qwen2.5-7B-Instruct2025.05 | 73.92 | — | — | — | |
| GRACEBackbone=Qwen2.5-7B-Instruct2025.05 | 73.92 | — | — | — | |
| DEDataset-free=false2026.02 | 73.08 | — | — | — | |
| DEBackbone=Qwen2.5-7B-Instruct2025.05 | 73.08 | — | — | — | |
| PromptWizardBackbone=Qwen2.5-7B-Instruct2025.05 | 73.05 | — | — | — | |
| SCModel=Qwen, k-shot=8, Seeds=52025.05 | 72.5 | — | — | — | |
| State VectorBackbone=Qwen-2.5-7B2026.05 | 70.12 | — | — | — | |
| APODataset-free=false2026.02 | 69.8 | — | — | — | |
| APOBackbone=Qwen2.5-7B-Instruct2025.05 | 69.8 | — | — | — | |
| NIDataset-free=true2026.02 | 68.1 | — | — | — | |
| NIBackbone=Qwen2.5-7B-Instruct2025.05 | 68.1 | — | — | — | |
| SCModel=Mistral, k-shot=8, Seeds=52025.05 | 67.5 | — | — | — | |
| Function VectorBackbone=Qwen-3-8B2026.05 | 67.2 | — | — | — | |
| LTVBackbone=LLaMA-3.1-8B2026.05 | 67.16 | — | — | — | |
| PRLBackbone=Qwen2.5-7B-Instruct, Prompt Selection=False2025.05 | 66.98 | — | — | — | |
| Zero-shotBackbone=Qwen-3-8B2026.05 | 66.4 | — | — | — | |
| BCModel=Llama, k-shot=8, Seeds=52025.05 | 65.78 | — | — | — | |
| GPS-SR-0.1Dataset-free=true2026.02 | 65.1 | — | — | — | |
| GPS-JDataset-free=true2026.02 | 64.2 | — | — | — | |
| SCModel=Llama, k-shot=8, Seeds=52025.05 | 63.05 | — | — | — | |
| State VectorBackbone=LLaMA-3.1-8B2026.05 | 62.52 | — | — | — | |
| I2CLBackbone=LLaMA-3.1-8B2026.05 | 62.28 | — | — | — | |
| Task VectorBackbone=LLaMA-3.1-8B2026.05 | 61.12 | — | — | — | |
| I2CLBackbone=Qwen-3-8B2026.05 | 59.76 | — | — | — | |
| Function VectorBackbone=LLaMA-3.1-8B2026.05 | 59.52 | — | — | — | |
| Zero-shotBackbone=LLaMA-3.1-8B2026.05 | 59.2 | — | — | — | |
| Function VectorBackbone=Qwen-2.5-7B2026.05 | 58.32 | — | — | — | |
| MIDataset-free=true2026.02 | 57.95 | — | — | — | |
| MIBackbone=Qwen2.5-7B-Instruct2025.05 | 57.95 | — | — | — | |
| Task VectorBackbone=Qwen-2.5-7B2026.05 | 54.96 | — | — | — | |
| Base LLMModel=Mistral, k-shot=8, Seeds=52025.05 | 54.84 | — | — | — | |
| ICLBackbone=LLaMA-2-7B2026.05 | 54.48 | — | — | — | |
| Zero-shotBackbone=LLaMA-2-7B2026.05 | 51.6 | — | — | — | |
| I2CLBackbone=LLaMA-2-7B2026.05 | 51.16 | — | — | — | |
| LTVBackbone=LLaMA-2-7B2026.05 | 50.32 | — | — | — | |
| State VectorBackbone=LLaMA-2-7B2026.05 | 50.2 | — | — | — | |
| Function VectorBackbone=LLaMA-2-7B2026.05 | 50.04 | — | — | — | |
| Function VectorBackbone=LLaMA-2-13B2026.05 | 50 | — | — | — | |
| I2CLBackbone=LLaMA-2-13B2026.05 | 50 | — | — | — | |
| Zero-shotBackbone=LLaMA-2-13B2026.05 | 49.8 | — | — | — | |
| Task VectorBackbone=LLaMA-2-13B2026.05 | 49.72 | — | — | — | |
| State VectorBackbone=LLaMA-2-13B2026.05 | 49.4 | — | — | — | |
| Zero-shotBackbone=Qwen-2.5-7B2026.05 | 48.8 | — | — | — | |
| Base LLMModel=Llama, k-shot=8, Seeds=52025.05 | 48.75 | — | — | — | |
| BCModel=Mistral, k-shot=8, Seeds=52025.05 | 48.75 | — | — | — | |
| Task VectorBackbone=LLaMA-2-7B2026.05 | 48.72 | — | — | — | |
| DCModel=Qwen, k-shot=8, Seeds=52025.05 | 47.66 | — | — | — | |
| CCModel=Qwen, k-shot=8, Seeds=52025.05 | 46.64 | — | — | — | |
| Base LLMModel=Qwen, k-shot=8, Seeds=52025.05 | 46.09 | — | — | — | |
| CCModel=Llama, k-shot=8, Seeds=52025.05 | 46.09 | — | — | — | |
| DCModel=Llama, k-shot=8, Seeds=52025.05 | 46.09 | — | — | — | |
| CCModel=Mistral, k-shot=8, Seeds=52025.05 | 46.09 | — | — | — | |
| DCModel=Mistral, k-shot=8, Seeds=52025.05 | 46.09 | — | — | — | |
| I2CLBackbone=Qwen-2.5-7B2026.05 | 42.72 | — | — | — | |
| STRIPAttack=EP, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.967 | 0.999 | 0.0475 | 0.991 | |
| ONIONAttack=EP, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.967 | 0.999 | 0.0525 | 0.041 | |
| RAPAttack=EP, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.967 | 0.999 | 0.047 | 0.3325 | |
| MDPAttack=EP, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.967 | 0.999 | 0.049 | 0.103 | |
| STRIPAttack=LWP, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.9615 | 0.991 | 0.0455 | 0.987 | |
| ONIONAttack=LWP, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.9615 | 0.991 | 0.0465 | 0.074 | |
| RAPAttack=LWP, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.9615 | 0.991 | 0.01 | 0.186 | |
| MDPAttack=LWP, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.9615 | 0.991 | 0.054 | 0.109 | |
| STRIPAttack=BadNets, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.9605 | 0.942 | 0.051 | 0.6885 | |
| ONIONAttack=BadNets, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.9605 | 0.942 | 0.035 | 0.166 | |
| RAPAttack=BadNets, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.9605 | 0.942 | 0.124 | 0.4365 | |
| MDPAttack=BadNets, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.9605 | 0.942 | 0.053 | 0.079 | |
| STRIPAttack=AddSent, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.959 | 0.97 | 0.025 | 0.855 | |
| ONIONAttack=AddSent, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.959 | 0.97 | 0.043 | 0.342 | |
| RAPAttack=AddSent, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.959 | 0.97 | 0.073 | 0.682 | |
| MDPAttack=AddSent, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.959 | 0.97 | 0.0485 | 0.09 | |
| STRIPAttack=SOS, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.949 | 0.996 | 0.0515 | 0.755 | |
| ONIONAttack=SOS, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.949 | 0.996 | 0.049 | 0.613 | |
| RAPAttack=SOS, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.949 | 0.996 | 0.001 | 0.291 | |
| MDPAttack=SOS, Learning Strategy=prompt-based fine-tuning, FRR threshold=5%2023.09 | 0.949 | 0.996 | 0.0535 | 0.041 |