System Prompt Extraction on System Prompts 41-model
0.915Self-Consistency Scoremicrosoft/phi-4
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| microsoft/phi-4ID=true, Prin.=true, Prio.=false, Cons.=true, Ref.=false, Method=L142026.01 | 0.915 | 0.782 | 0.849 | |
| openai/o1ID=true, Prin.=true, Prio.=true, Cons.=true, Ref.=false, Method=H8+H42026.01 | 0.909 | 0.871 | 0.89 | |
| perplexity/sonar-proID=true, Prin.=true, Prio.=false, Cons.=true, Ref.=false, Method=H42026.01 | 0.905 | 0.79 | 0.848 | |
| deepseek/deepseek-v3.2ID=true, Prin.=true, Prio.=true, Cons.=true, Ref.=true, Method=H8+H42026.01 | 0.903 | 0.875 | 0.889 | |
| meta-llama/llama-4-mav.ID=true, Prin.=true, Prio.=false, Cons.=true, Ref.=false, Method=L14+L72026.01 | 0.901 | 0.857 | 0.879 | |
| anthropic/claude-opus-4.5ID=true, Prin.=true, Prio.=true, Cons.=true, Ref.=true, Method=H8+H42026.01 | 0.891 | 0.856 | 0.874 | |
| google/gemini-3-proID=true, Prin.=true, Prio.=true, Cons.=true, Ref.=false, Method=H8+H42026.01 | 0.867 | 0.832 | 0.85 | |
| cohere/command-aID=true, Prin.=true, Prio.=false, Cons.=true, Ref.=true, Method=L14+L72026.01 | 0.865 | 0.721 | 0.793 | |
| openai/gpt-5.2-codexID=true, Prin.=true, Prio.=true, Cons.=true, Ref.=true, Method=H8 (11-turn)2026.01 | 0.846 | 0.809 | 0.828 | |
| tencent/hunyuan-a13bID=true, Prin.=true, Prio.=false, Cons.=true, Ref.=false, Method=L14+L72026.01 | 0.821 | 0.82 | 0.821 |