Share your thoughts, 1 month free Claude Pro on usSee more

Prompt Injection Detection on PopUp attack (top visited websites)

99.98Accuracy

WARD-0.8b

Updated 2mo ago

Evaluation Results

Method
WARD-0.8b 2026.05	99.98	100	99.96	99.98
WARD-2b 2026.05	99.91	100	99.81	99.91
Claude-Sonnet-4.6 2026.05	99.66	99.7	99.63	99.66
Gemini-3-Flash 2026.05	98.96	99.78	98.17	98.96
GPT-5.4 2026.05	98.9	99.59	98.23	98.91
GPT-5.4-Mini 2026.05	98.21	96.68	99.73	98.18
Claude-Haiku-4.5 2026.05	97.86	98.06	97.66	97.86
WebAgentGuard-8B 2026.04	91.13	90.18	91.31	90.74
Gemini-3.1-pro 2026.05	91.03	99.48	85.1	91.73
GPT-4.1 2026.04	90.45	80.08	100	88.94
Qwen-3.5-4B 2026.05	88.09	77.18	98.71	86.63
Gemma-4-31B 2026.05	86.99	74.72	99.01	85.17
Qwen3.5-35B-A3B 2026.05	86.89	75.58	97.69	85.22
WebAgentGuard-8b 2026.05	86.63	76.14	96.37	85.07
Qwen-3.5-2B 2026.05	84.86	84.65	85.02	84.83
GPT-4o 2026.04	82.78	78.86	99.86	88.13
WebAgentGuard-4B 2026.04	82.11	70.55	89.98	79.09
Qwen-3.5-0.8B 2026.05	76.34	74.31	77.46	75.85
Qwen3-VL-Instruct-4B 2026.04	72.21	44.37	97.03	60.9
GPT-4o-Mini 2026.04	70.06	41.52	91.78	57.18
WebAgentGuard-4b 2026.05	66.93	37.55	91.05	53.17
Gemma-3-4B 2026.05	65.6	68.94	64.63	66.71
Qwen3-VL-Instruct-8B 2026.04	62.8	22.62	99.09	36.83
Qwen2.5-VL-Instruct-7B 2026.04	59.53	21.35	87.44	34.32
Prompt-Guard-2-86M 2026.04	52.05	2.06	50	3.96
GuardReasoner-VL-7B 2026.04	52	1.34	50	2.61
Llama-Guard-3-Vision-11B 2026.04	51.65	4.02	45.34	7.39
Prompt-Guard-1-86M 2026.04	50.81	33.12	48.12	39.24
Llama-Guard-4-12B 2026.05	50.04	0.26	58.33	0.52
GPT-Oss-Safeguard-20b 2026.05	50	0	0	0
Prompt-Guard-1-86M 2026.05	50	1.79	50	3.46
Prompt-Guard-2-86M 2026.05	50	0.3	50	0.59
Llama-Guard-3-Vision-11B 2026.05	50	0	0	0
GuardReasoner-VL-7B 2026.05	50	0	0	0
BrowseSafe 2026.05	50	0	0	0
SuperAgent-Guard-1.7b 2026.05	50	0	0	0
SuperAgent-Guard-4b 2026.05	50	0	0	0
DataSentinel 2026.05	49.96	48.77	49.96	49.36
PromptArmor 2026.05	49.78	87.92	49.87	63.64
Llama-3.2-Vision-Instruct-11B 2026.04	44.92	5.53	40.87	9.74