Uncertainty Quantification on Vision Datasets averaged (test)
81.7AUROCSGPU
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SGPULVLM Architecture=llava-mistral-7b2025.12 | 81.7 | 82.2 | 27.4 | |
| SGPULVLM Architecture=Qwen2.5-VL-3B2025.12 | 81 | 80.5 | 10.2 | |
| UMPIRELVLM Architecture=llava-mistral-7b2025.12 | 78.2 | 68.8 | 19.5 | |
| SGPULVLM Architecture=Qwen2.5-VL-7B2025.12 | 76.6 | 76.7 | 8 | |
| Cov EigenscoreLVLM Architecture=idefics2-8b2025.12 | 76.5 | 67.8 | 22.9 | |
| SELVLM Architecture=llava-mistral-7b2025.12 | 75 | 69.1 | 18.6 | |
| DSELVLM Architecture=llava-mistral-7b2025.12 | 75 | 67.8 | 18.9 | |
| ManualMaternLVLM Architecture=Qwen2.5-VL-3B2025.12 | 74.7 | 69.3 | 14.5 | |
| SGPULVLM Architecture=idefics2-8b2025.12 | 74.6 | 75.3 | 24.2 | |
| Cos EigenscoreLVLM Architecture=llava-mistral-7b2025.12 | 73.7 | 69.4 | 18 | |
| PELVLM Architecture=llava-mistral-7b2025.12 | 73.6 | 68.5 | 12.7 | |
| Cos EigenscoreLVLM Architecture=idefics2-8b2025.12 | 73.6 | 72.6 | 26.6 | |
| SELVLM Architecture=Qwen2.5-VL-3B2025.12 | 73.1 | 68.3 | 16.5 | |
| DSELVLM Architecture=Qwen2.5-VL-3B2025.12 | 73 | 68.1 | 15.4 | |
| PELVLM Architecture=Qwen2.5-VL-3B2025.12 | 72.6 | 68.2 | 8.7 | |
| ManualHeatLVLM Architecture=Qwen2.5-VL-3B2025.12 | 72.6 | 68.4 | 16.4 | |
| ManualMaternLVLM Architecture=Qwen2.5-VL-7B2025.12 | 72.4 | 69.8 | 9.7 | |
| ManualMaternLVLM Architecture=llava-mistral-7b2025.12 | 72.4 | 69.7 | 8.9 | |
| UMPIRELVLM Architecture=Qwen2.5-VL-3B2025.12 | 72.2 | 61.8 | 15.8 | |
| Cov EigenscoreLVLM Architecture=Qwen2.5-VL-7B2025.12 | 71.7 | 63.1 | 17.7 | |
| ManualHeatLVLM Architecture=Qwen2.5-VL-7B2025.12 | 71.5 | 69.2 | 20.7 | |
| UMPIRELVLM Architecture=idefics2-8b2025.12 | 71.4 | 72.7 | 28.5 | |
| Cos EigenscoreLVLM Architecture=Qwen2.5-VL-7B2025.12 | 71.4 | 62.5 | 15.8 | |
| SELVLM Architecture=Qwen2.5-VL-7B2025.12 | 71.3 | 69 | 24.4 | |
| PELVLM Architecture=Qwen2.5-VL-7B2025.12 | 71.3 | 69.2 | 10.7 | |
| DSELVLM Architecture=Qwen2.5-VL-7B2025.12 | 71.2 | 68.7 | 24.3 | |
| UMPIRELVLM Architecture=Qwen2.5-VL-7B2025.12 | 70.9 | 65.8 | 27.8 | |
| Cov EigenscoreLVLM Architecture=llava-mistral-7b2025.12 | 70.9 | 67.9 | 28.8 | |
| PELVLM Architecture=idefics2-8b2025.12 | 70.2 | 72.4 | 25.2 | |
| Cos EigenscoreLVLM Architecture=Qwen2.5-VL-3B2025.12 | 69.9 | 55.9 | 15.7 | |
| SELVLM Architecture=idefics2-8b2025.12 | 69.8 | 72 | 25 | |
| ManualMaternLVLM Architecture=idefics2-8b2025.12 | 69.3 | 71.5 | 23 | |
| ManualHeatLVLM Architecture=idefics2-8b2025.12 | 68.8 | 71.4 | 24.9 | |
| ManualHeatLVLM Architecture=llava-mistral-7b2025.12 | 68.7 | 69.4 | 23.3 | |
| DSELVLM Architecture=idefics2-8b2025.12 | 67.2 | 71.6 | 24.3 | |
| Cov EigenscoreLVLM Architecture=Qwen2.5-VL-3B2025.12 | 65.3 | 55.4 | 19.6 |