MM Alignment on HallusionBench
67.9ScoreYuvion VL-32B
Evaluation Results
| Method | Links | |
|---|---|---|
| Yuvion VL-32BDecoding Mode=Reasoning, Model Scale=32B2026.06 | 67.9 | |
| Qwen3-VL-8BDecoding Mode=Thinking, Model Scale=8B2026.06 | 65.4 | |
| Yuvion VL-8BDecoding Mode=Reasoning, Model Scale=8B2026.06 | 63.6 | |
| Qwen3.5-PlusParameter Count=397B2026.06 | 57.1 | |
| Qwen3-VL-32BDecoding Mode=Thinking, Model Scale=32B2026.06 | 56.3 | |
| Qwen3-VL-32BDecoding Mode=Instruct, Model Scale=32B2026.06 | 55.6 | |
| K2.5Parameter Count=≈1T2026.06 | 53.2 | |
| Qwen3-VL-8BDecoding Mode=Instruct, Model Scale=8B2026.06 | 50.6 | |
| Yuvion VL-32BDecoding Mode=Instruct, Model Scale=32B2026.06 | 50.1 | |
| GPT-5.4Parameter Count=≈1.5T2026.06 | 49.5 | |
| Yuvion VL-8BDecoding Mode=Instruct, Model Scale=8B2026.06 | 46.8 | |
| Opus-4.6Parameter Count=≈800B2026.06 | 46.6 |