Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Secret Data Extraction on 100 different secrets
Loading...
100
Success Rate (P > 0.10)
LLP
90.64
93.07
95.5
97.93
Jun 15, 2026
Success Rate (P > 0.10)
Success Rate (P > 0.30)
Success Rate (P > 0.50)
Updated 1mo ago
Evaluation Results
Method
Method
Links
Success Rate (P > 0.10)
Success Rate (P > 0.30)
Success Rate (P > 0.50)
LLP
LLM=OPT (250M)
2026.06
100
83
70
LLP
LLM=GPT-2 Medium (355M)
2026.06
100
91
76
LLP
LLM=Llama2 (7B)
2026.06
100
100
100
LLP
LLM=Llama2 (13B)
2026.06
100
100
98
LLP
LLM=Llama3.2 (1B)
2026.06
100
100
100
LLP
LLM=Pythia (160M)
2026.06
98
79
64
Llava1.5
Parameters=7B
2026.06
98
97
90
InstructBlip
Parameters=7B
2026.06
96
94
88
LLP
LLM=GPT-Neo (125M)
2026.06
95
76
60
LLP
LLM=GPT-2 Small (124M)
2026.06
94
74
57
LLP
LLM=DistilGPT-2 (82M)
2026.06
92
68
54
InstructBlip
Parameters=4B
2026.06
91
90
83
Feedback
Search any
task
Search any
task
