Instruction Following on Alpaca clean (test)
79.81F1 ScoreBadNets
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BadNetsLLM Architecture=Llama 2 7B2026.05 | 79.81 | 25.95 | |
| SynsLLM Architecture=Mistral 7B2026.05 | 79 | 23.6 | |
| SyndLLM Architecture=Mistral 7B2026.05 | 74.12 | 18.15 | |
| SleeperLLM Architecture=Mistral 7B2026.05 | 74.05 | 18.4 | |
| BadNetsLLM Architecture=Mistral 7B2026.05 | 74.01 | 18.5 | |
| VPILLM Architecture=Mistral 7B2026.05 | 73.99 | 18.6 | |
| SyndLLM Architecture=Llama 2 7B2026.05 | 73.78 | 18.45 | |
| SleeperLLM Architecture=Llama 2 7B2026.05 | 73.72 | 18.95 | |
| CBALLM Architecture=Llama 2 7B2026.05 | 73.7 | 18.91 | |
| SyndLLM Architecture=Vicuna 7B2026.05 | 73.65 | 18.4 | |
| VPILLM Architecture=Vicuna 7B2026.05 | 73.62 | 19.15 | |
| VPILLM Architecture=Llama 2 7B2026.05 | 73.58 | 18.85 | |
| BaseLLM Architecture=Vicuna 7B2026.05 | 73.57 | 19.25 | |
| SleeperLLM Architecture=Vicuna 7B2026.05 | 73.51 | 19.1 | |
| CBALLM Architecture=Mistral 7B2026.05 | 73.51 | 18.96 | |
| BaseLLM Architecture=Llama 2 7B2026.05 | 73.49 | 18.4 | |
| BadNetsLLM Architecture=Vicuna 7B2026.05 | 73.4 | 18.4 | |
| BaseLLM Architecture=Mistral 7B2026.05 | 73.39 | 17.7 | |
| ParaesthesiaLLM Architecture=Llama 2 7B2026.05 | 73.35 | 18 | |
| SynsLLM Architecture=Llama 2 7B2026.05 | 73.32 | 18.45 | |
| ParaesthesiaLLM Architecture=Mistral 7B2026.05 | 73.28 | 18.4 | |
| ParaesthesiaLLM Architecture=Vicuna 7B2026.05 | 73.27 | 18.15 | |
| SynsLLM Architecture=Vicuna 7B2026.05 | 73.12 | 18 | |
| CBALLM Architecture=Vicuna 7B2026.05 | 73.06 | 17.96 | |
| CBALLM Architecture=Qwen2.5 7B Instruct2026.05 | 72.81 | 16.17 | |
| VPILLM Architecture=Qwen2.5 7B Instruct2026.05 | 72.62 | 15.4 | |
| SyndLLM Architecture=Qwen2.5 7B Instruct2026.05 | 72.61 | 16 | |
| SleeperLLM Architecture=Qwen2.5 7B Instruct2026.05 | 72.57 | 15.2 | |
| ParaesthesiaLLM Architecture=Qwen2.5 7B Instruct2026.05 | 72.57 | 15.95 | |
| BaseLLM Architecture=Qwen2.5 7B Instruct2026.05 | 72.54 | 16.3 | |
| BadNetsLLM Architecture=Qwen2.5 7B Instruct2026.05 | 72.51 | 15.2 | |
| SynsLLM Architecture=Qwen2.5 7B Instruct2026.05 | 72.44 | 15.8 |