Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Utility Evaluation on GPQA Diamond
Loading...
53
Accuracy (pass@1)
No Defense
18.68
27.59
36.5
45.41
Aug 6, 2025
Accuracy (pass@1)
Updated 27d ago
Evaluation Results
Method
Method
Links
Accuracy (pass@1)
No Defense
Backbone=R1-Qwen-7B
2025.08
53
SAFEPATH-FT
Backbone=R1-Qwen-7B
2025.08
53
ReasoningGuard
Backbone=R1-Qwen-7B
2025.08
53
SafeDecoding
Backbone=R1-Qwen-7B
2025.08
52
Self-Reminder
Backbone=R1-Qwen-7B
2025.08
52
No Defense
Backbone=R1-Llama-8B
2025.08
52
ThinkingI
Backbone=R1-Qwen-7B
2025.08
50
SafeKey
Backbone=R1-Qwen-7B
2025.08
49
SafeDecoding
Backbone=R1-Llama-8B
2025.08
48
RealSafe-R1
Backbone=R1-Qwen-7B
2025.08
46
SAFEPATH-ZS
Backbone=R1-Qwen-7B
2025.08
46
SAFEPATH-FT
Backbone=R1-Llama-8B
2025.08
45
SAFEPATH-ZS
Backbone=R1-Llama-8B
2025.08
45
RealSafe-R1
Backbone=R1-Llama-8B
2025.08
44
Self-Reminder
Backbone=R1-Llama-8B
2025.08
44
ReasoningGuard
Backbone=R1-Llama-8B
2025.08
44
SafeKey
Backbone=R1-Llama-8B
2025.08
41
ThinkingI
Backbone=R1-Llama-8B
2025.08
41
SmoothLLM
Backbone=R1-Qwen-7B
2025.08
34
SmoothLLM
Backbone=R1-Llama-8B
2025.08
31
Paraphrase
Backbone=R1-Qwen-7B
2025.08
23
Paraphrase
Backbone=R1-Llama-8B
2025.08
20
Feedback
Search any
task
Search any
task