Visual Question Answering on MMA-Bench (Adversarial Robustness)
96.79Clean AccuracyGemini-2.5-Flash-Lite
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Gemini-2.5-Flash-LiteModel=Gemini-2.5-Flash-Lite2026.04 | 96.79 | 93.1 | 0 | 3.81 | |
| Gemini-3.1-Flash-Lite-previewModel=Gemini-3.1-Flash-Lite-preview2026.04 | 96.58 | 93.16 | 0 | 3.79 | |
| Qwen3-Omni-30BModel=Qwen3-Omni-30B2026.04 | 92.88 | 86.93 | 0 | 5.17 | |
| Qwen2.5-Omni-7BModel=Qwen2.5-Omni-7B2026.04 | 76.68 | 63.83 | 0 | 24.27 | |
| ChatBridgeModel=ChatBridge2026.04 | 51.64 | 44.13 | 0 | 5.1 | |
| PandaGPTModel=PandaGPT2026.04 | 28.75 | 18.54 | 0 | 0.76 |