List your company

ResearchFormerly WizWand

WildGuard

Benchmarks

TaskDatasetSOTA resultTrendResults

Harmfulness Detection

WildGuard

90.6Macro F1 Score

47

Safety classification

WildGuard (test)

88.8F1 Score

35

Response Harmfulness Classification

WildGuard (test)

79.48F1 (Total)

30

Safety Evaluation

Wildguard (test)

0.08Wildguard Test Score

27

Text-based safety moderation

WildGuard

80.26F1 Score

26

Value Alignment

WildGuard (test)

29.69Score

24

Input Moderation

WildGuard (test)

89.5F1 Score

22

Prompt Harmfulness Classification

WILDGUARD (test)

89.44F1 Score

18

Safety Evaluation

WildGuard

25.5WildGuard Refusal Rate

16

Prompt Classification

WildGuard Text Prompt

90.46F1 Score

14

Refusal Detection

WILDGUARD (test)

94F1 (Harmful)

14

Binary safety classification

wildguard prompt safety

97.91Macro F1

11

Output Moderation

WildGuard (test)

79.5F1 Score

11

False Positive Rate Detection on Benign Prompts

WildGuard benign

3.44FPR

10

Stealthiness Evaluation

WildGuard 7B

2.33Mean Perplexity

10

Streaming Safety Detection

WildGuard (test)

83.45Det@1

8

Jailbreak Attack

WildGuard (test)

82.64ASR

8

Violation Detection

WildGuard (test)

89.17Safety F1

7

Refusal Evaluation

WildGuard Harmful

84.35Refusal Rate

7

Over-refusal Evaluation

WildGuard Unharmful

1.06Over-refusal Rate

7

Safety Moderation

WildGuard Prompt

89.5F1 Score

7

Audio Safety Moderation

WildGuard-TTS

88.4F1 Score

7

Safety Evaluation

WildGuard (Unsafe)

54.9Refusal Rate

6

Safety Evaluation

WildGuard (Safe)

30.7Refusal Rate

6

Multi-label Safety Categorization

wildguard prompt subcategory

83.35Macro Accuracy

4