Audio Alignment Detection on AnyAudio-Judge Chinese Subset
93.33Accuracy (Speech-Real)AnyAudio-Judge
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| AnyAudio-JudgePrompt=dynamic rubric2026.06 | 93.33 | 80.15 | 77.9 | 82.5 | 92.22 | 80.13 | 90.6 | 85.26 | |
| Gemini-2.5-ProPrompt=holistic2026.06 | 92.33 | 81.4 | 69.42 | 76.4 | 85.63 | 81.47 | 73.4 | 80.01 | |
| Qwen3-Omni-30B-A3B-InstructPrompt=dynamic rubric2026.06 | 90.75 | 78.65 | 71.6 | 77.33 | 76.25 | 75.88 | 67.3 | 76.82 | |
| Gemini-2.5-ProPrompt=dynamic rubric2026.06 | 90.67 | 80 | 72 | 77 | 79.67 | 73.33 | 75.5 | 78.31 | |
| Qwen3-Omni-30B-A3B-CaptionerPrompt=dynamic rubric2026.06 | 90.17 | 79.5 | 70.8 | 76.42 | 74.86 | 75.75 | 69.1 | 76.66 | |
| Kimi-Audio-7B-InstructPrompt=dynamic rubric2026.06 | 81.17 | 70.1 | 66.8 | 75.17 | 69.72 | 68 | 64.9 | 70.84 | |
| Qwen2.5-Omni-7BPrompt=dynamic rubric2026.06 | 78.17 | 72.1 | 68 | 75.58 | 74.72 | 71.63 | 63.3 | 71.93 | |
| Qwen3-Omni-30B-A3B-CaptionerPrompt=holistic2026.06 | 75.42 | 64.75 | 59.2 | 64 | 68.61 | 70 | 55.3 | 65.33 | |
| Qwen3-Omni-30B-A3B-InstructPrompt=holistic2026.06 | 67.33 | 58.1 | 54.1 | 57.92 | 64.58 | 66.75 | 52.9 | 60.24 | |
| MiDashengLMPrompt=dynamic rubric2026.06 | 66.94 | 65.5 | 67.9 | 75.83 | 71.25 | 71.13 | 59.9 | 68.35 | |
| Audio-Flamingo3Prompt=dynamic rubric2026.06 | 60.25 | 58.9 | 65.17 | 68.92 | 70.28 | 67.37 | 56.5 | 63.91 | |
| MiDashengLMPrompt=holistic2026.06 | 51.58 | 52.5 | 53.4 | 49.5 | 52.22 | 48.88 | 47.2 | 50.75 | |
| Kimi-Audio-7B-InstructPrompt=holistic2026.06 | 51 | 50.2 | 52.5 | 50.25 | 52.08 | 51.5 | 49 | 50.93 | |
| Audio-Flamingo3Prompt=holistic2026.06 | 50.25 | 50.35 | 65.9 | 67.66 | 59.97 | 59.52 | 52.11 | 57.97 | |
| Qwen2.5-Omni-7BPrompt=holistic2026.06 | 50.25 | 50.2 | 50.4 | 51.25 | 56.25 | 53.88 | 50.1 | 51.76 |