Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Over-Refusal Evaluation on OverBench
Loading...
5.59
ORR
w/o Safety-Aligned Optimization
4.5584
11.5217
18.485
25.4483
Jun 11, 2026
ORR
Updated 1mo ago
Evaluation Results
Method
Method
Links
ORR
w/o Safety-Aligned Optimization
Guardrail Type=Stream,...
2026.06
5.59
FreoStream
Guardrail Type=Stream,...
2026.06
6.64
Qwen-Stream-8B
Guardrail Type=Stream,...
2026.06
10.16
YuFeng-XGuard-0.6B Prompt+Response
Guardrail Type=Generat...
2026.06
12.69
GuardReasoner-1B
Guardrail Type=Generat...
2026.06
14.5
Qwen-Gen-0.6B Prompt Only
Guardrail Type=Generat...
2026.06
15.86
Qwen-Gen-0.6B Prompt+Response
Guardrail Type=Generat...
2026.06
16.28
YuFeng-XGuard-0.6B Prompt Only
Guardrail Type=Generat...
2026.06
26.17
Nemotron-4B
Guardrail Type=Generat...
2026.06
29.2
w/o Future-Aware Reasoning
Guardrail Type=Stream,...
2026.06
30.99
Qwen-Stream-0.6B
Guardrail Type=Stream,...
2026.06
31.38
Feedback
Search any
task
Search any
task