Instruction Induction on Instruction Induction (test)
100Active to PassiveAPE
Evaluation results
| Method | Links | |||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| APE2024.03 | 100 | 63.7 | 25 | 29.2 | 57.3 | 6.9 | 67.3 | 100 | 57.4 | 89.7 | 100 | 75.3 | 99.7 | 36.3 | 63.3 | 45.7 | 92.7 | 15.7 | 74.7 | 0 | 91.3 | 100 | 67.3 | 36 | 34.7 | 84 | 87 | 88.7 | 33 | 44 | 4 | 0.37 | — | — | — | — | — | — | — | — | — | |
| EvoPrompt2024.03 | 100 | 84 | 31 | 33 | 84 | 11.1 | 27.3 | 100 | 51.6 | 87.3 | 100 | 86 | 100 | 55 | 10 | 15 | 98 | 59.7 | 24.7 | 2 | 93 | 100 | 100 | 40.3 | 83 | 85 | 82.3 | 80.7 | 48 | 51.3 | 7 | 0.47 | — | — | — | — | — | — | — | — | — | |
| ZOPO2024.03 | 100 | 85.2 | 32.7 | 41.7 | 94.7 | 23.5 | 100 | 100 | 61.3 | 92.3 | 100 | 86.3 | 100 | 52.3 | 32 | 56.5 | 100 | 100 | 25.7 | 7.6 | 93.5 | 100 | 100 | 43.3 | 90 | 85.3 | 85.3 | 91 | 60 | 59.3 | 18 | 0.87 | — | — | — | — | — | — | — | — | — | |
| ZOPOGPT2024.03 | 100 | 84 | 27 | 29.2 | 80 | 2.8 | 100 | 100 | 61.9 | 92.7 | 100 | 77.7 | 100 | 40.3 | 68.7 | 71 | 94.7 | 61 | 96.7 | 37.3 | 89.3 | 100 | 100 | 44.7 | 92.3 | 84.7 | 86.3 | 86.7 | 60.3 | 58.3 | 15 | 0.73 | — | — | — | — | — | — | — | — | — | |
| Worst Human-Written PromptBase LLM=gpt-3.5-turbo-0301, Number of input prompts=1, Number of output candidate prompts=8, Number of iteration rounds=12023.08 | 100 | 77 | — | — | 0 | 5 | 87 | 60 | 46 | 20 | 56 | 50 | 44 | — | — | 35 | — | 0 | 95 | 0 | 52 | 99 | 7 | 11 | 42 | 81 | 87 | 88 | — | — | — | — | 0 | 47 | — | — | — | — | — | — | — | |
| PROmptingBase LLM=gpt-3.5-turbo-0301, Number of input prompts=1, Number of output candidate prompts=8, Number of iteration rounds=12023.08 | 100 | 82 | — | — | 0 | 8 | 100 | 0 | 53 | 59 | 48 | 78 | 100 | — | — | 37 | — | 0.02 | 45 | 0.05 | 13 | 98 | 99 | 15 | 73 | 82 | 86 | 86 | — | — | — | — | 0 | 53 | — | — | — | — | — | — | — | |
| EvoPromptBase LLM=gpt-3.5-turbo-0301, Number of input prompts=1, Number of output candidate prompts=8, Number of iteration rounds=12023.08 | 100 | 82 | — | — | 0 | 6 | 100 | 20 | 53 | 7 | 46 | 23 | 12 | — | — | 43 | — | 0 | 45 | 0.05 | 22 | 97 | 100 | 14 | 73 | 82 | 83 | 87 | — | — | — | — | 0.01 | 46 | — | — | — | — | — | — | — | |
| InstructZero2024.03 | 99.7 | 82.7 | 25.7 | 37.5 | 81.33 | 8.6 | 69.3 | 100 | 53.1 | 90 | 59 | 77.7 | 100 | 36 | 61.3 | 50.7 | 86.7 | 100 | 43.3 | 0 | 87.7 | 98.7 | 100 | 27.7 | 71.7 | 82.3 | 87.3 | 87.7 | 31 | 59 | 4 | 0.43 | — | — | — | — | — | — | — | — | — | |
| PACEBase LLM=gpt-3.5-turbo-0301, Number of input prompts=1, Number of output candidate prompts=8, Number of iteration rounds=12023.08 | 99 | 85 | — | — | 0.53 | 6 | 100 | 100 | 59 | 53 | 100 | 76 | 100 | — | — | 44 | — | 0.56 | 100 | 0.42 | 91 | 100 | 100 | 12 | 92 | 84 | 83 | 88 | — | — | — | — | 0.16 | 72 | — | — | — | — | — | — | — | |
| INSTINCT2024.03 | 97 | 84.7 | 25 | 29.2 | 58.7 | 21.3 | 100 | 93 | 55.3 | 93.7 | 100 | 81.7 | 100 | 34 | 70 | 66.7 | 92.7 | 100 | 10 | 14 | 89.7 | 100 | 100 | 30.7 | 85.7 | 84 | 88 | 83 | 51.3 | 63.3 | 10 | 0.57 | — | — | — | — | — | — | — | — | — | |
| APEPrompt-LM=Mistral-7B, Target-LM=Gemma-7B, Algorithm=Automatic Prompt Engineer2026.02 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 51.94 | 43.81 | 67.19 | 4,450 | 35.88 | 45.92 | 7,469 | — | |
| Few-shotPrompt-LM=Mistral-7B, Target-LM=Gemma-7B, Mode=Few-shot2026.02 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 5.21 | 3.75 | 0 | 60 | 0 | 35 | 50 | — | |
| GFLOWPOPrompt-LM=Mistral-7B, Target-LM=Gemma-7B, Algorithm=GFLOWPO2026.02 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 64.96 | 59.17 | 73.25 | 5,753 | 52.58 | 63.78 | 8,500 | — | |
| ManualPrompt-LM=Mistral-7B, Target-LM=Gemma-7B, Mode=Manual prompting2026.02 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 33.7 | 29.62 | 36.75 | 1,460 | 23.25 | 49.67 | 5,710 | — | |
| ProTEGIPrompt-LM=Mistral-7B, Target-LM=Gemma-7B, Algorithm=ProTEGI2026.02 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 51.6 | 43.75 | 68.75 | 2,750 | 56.25 | 58.33 | 6,230 | — | |
| StablePromptPrompt-LM=Mistral-7B, Target-LM=Gemma-7B, Algorithm=StablePrompt2026.02 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | 57.71 | 50.58 | 70.83 | 4,046 | 47 | 59.67 | 8,255 | — |