Gemini
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
Gemini Flash 2.5 (test)
0ASR
27
Gemini Flash (Small Target) 2.5
98Similarity Score
18
Gemini 1.5-pro 2.5-flash (test)
67.4ASR
16
Gemini 2.0
41.3ASR
11
Gemini-2.0 Flash generated text
99.34AUROC (Insertion)
10
Gemini 2.5-Pro
0.87KMRa
9
Gemini Pro 3
92.5ASR
9
Gemini-3-flash closed-source standard MLLMs
4Attack Success Rate (ASR)
8
Gemini-1.5 3-flash (test)
50.8ASR
8
Gemini 3.1
70.2Attack Success Rate (ASR)
8
Gemini 2.5
81.3ASR
8
Gemini Image Captioning Hard Criterion 1.5
81ASR
8
Gemini 100 multi-shot video prompts 2.5 Pro
0.825Intra-shot Consistency (Subject)
8
Gemini Flash 2.0 (test)
95.24Accuracy
7
Gemini-3 generated text
92.84AUROC
7
Gemini flash 1.5
81.33Detoxify Score
5
Gemini-2-Flash
3.65Compliance
5
Gemini-3-flash
51ASR
4
Gemini flash 1.5
83KMR (alpha)
4
Gemini Flash 3
90.5Attack Success Rate
4
Gemini Flash 1.5
58Attack Success Rate (T1)
4
Gemini 2.0
520ASR
4
Gemini Targeted Task Pairs
0.179LPIPS
2
Gemini Targeted Task Pairs
10.6CIEDE
2
Gemini Targeted Task Pairs
18.56CIEDE
2