Medical Visual Question Answering on VQA-Med Closed (AUROC)
81.8AUROCHAC-Platt
Evaluation Results
| Method | Links | |
|---|---|---|
| HAC-PlattBase Model=InternVL3-38B, Confidence Type=Verbalized2026.04 | 81.8 | |
| HAC-GateBase Model=InternVL3-38B, Confidence Type=Verbalized2026.04 | 81.8 | |
| Hist. BinningBase Model=InternVL3-2B, Confidence Type=Sampling2026.04 | 79.9 | |
| HAC-PlattBase Model=InternVL3-8B, Confidence Type=Sampling2026.04 | 79.3 | |
| HAC-GateBase Model=InternVL3-8B, Confidence Type=Sampling2026.04 | 79.3 | |
| UncalibratedBase Model=InternVL3-38B, Confidence Type=Verbalized2026.04 | 79.1 | |
| Platt ScalingBase Model=InternVL3-38B, Confidence Type=Verbalized2026.04 | 79.1 | |
| HAC-PlattBase Model=Qwen3-VL-8B, Confidence Type=Sampling2026.04 | 78.6 | |
| HAC-PlattBase Model=Qwen3-VL-8B, Confidence Type=Verbalized2026.04 | 78 | |
| Isotonic Regr.Base Model=InternVL3-38B, Confidence Type=Verbalized2026.04 | 77.7 | |
| HAC-GateBase Model=Qwen3-VL-8B, Confidence Type=Verbalized2026.04 | 77.3 | |
| HAC-GateBase Model=Qwen3-VL-8B, Confidence Type=Sampling2026.04 | 77.1 | |
| UncalibratedBase Model=InternVL3-8B, Confidence Type=Sampling2026.04 | 76.2 | |
| Platt ScalingBase Model=InternVL3-8B, Confidence Type=Sampling2026.04 | 76.2 | |
| Isotonic Regr.Base Model=InternVL3-2B, Confidence Type=Sampling2026.04 | 75.6 | |
| UncalibratedBase Model=InternVL3-2B, Confidence Type=Sampling2026.04 | 73.9 | |
| Platt ScalingBase Model=InternVL3-2B, Confidence Type=Sampling2026.04 | 73.9 | |
| Hist. BinningBase Model=InternVL3-38B, Confidence Type=Verbalized2026.04 | 73.6 | |
| HAC-PlattBase Model=InternVL3-2B, Confidence Type=Sampling2026.04 | 73.3 | |
| HAC-GateBase Model=InternVL3-2B, Confidence Type=Sampling2026.04 | 73.3 | |
| Isotonic Regr.Base Model=InternVL3-8B, Confidence Type=Sampling2026.04 | 72.5 | |
| Hist. BinningBase Model=InternVL3-8B, Confidence Type=Sampling2026.04 | 65.3 | |
| HAC-PlattBase Model=Qwen3-VL-32B, Confidence Type=Verbalized2026.04 | 64.7 | |
| HAC-PlattBase Model=InternVL3-8B, Confidence Type=Verbalized2026.04 | 64.1 | |
| HAC-GateBase Model=LLaVA-NeXT-7B, Confidence Type=Verbalized2026.04 | 63.7 | |
| HAC-GateBase Model=Qwen3-VL-32B, Confidence Type=Verbalized2026.04 | 63.3 | |
| HAC-PlattBase Model=LLaVA-NeXT-7B, Confidence Type=Verbalized2026.04 | 63.1 | |
| HAC-GateBase Model=InternVL3-8B, Confidence Type=Verbalized2026.04 | 62.3 | |
| UncalibratedBase Model=Qwen3-VL-32B, Confidence Type=Verbalized2026.04 | 60.7 | |
| Platt ScalingBase Model=Qwen3-VL-32B, Confidence Type=Verbalized2026.04 | 60.7 | |
| UncalibratedBase Model=InternVL3-2B, Confidence Type=Verbalized2026.04 | 58.7 | |
| Platt ScalingBase Model=InternVL3-2B, Confidence Type=Verbalized2026.04 | 58.7 | |
| HAC-PlattBase Model=InternVL3-2B, Confidence Type=Verbalized2026.04 | 58.7 | |
| Hist. BinningBase Model=InternVL3-38B, Confidence Type=Sampling2026.04 | 58.7 | |
| HAC-GateBase Model=InternVL3-2B, Confidence Type=Verbalized2026.04 | 57.4 | |
| HAC-GateBase Model=LLaVA-NeXT-34B, Confidence Type=Sampling2026.04 | 57.1 | |
| UncalibratedBase Model=LLaVA-NeXT-34B, Confidence Type=Sampling2026.04 | 56.9 | |
| Platt ScalingBase Model=LLaVA-NeXT-34B, Confidence Type=Sampling2026.04 | 56.9 | |
| UncalibratedBase Model=Qwen3-VL-2B, Confidence Type=Sampling2026.04 | 56.7 | |
| Platt ScalingBase Model=Qwen3-VL-2B, Confidence Type=Sampling2026.04 | 56.7 | |
| Isotonic Regr.Base Model=Qwen3-VL-2B, Confidence Type=Sampling2026.04 | 56.7 | |
| UncalibratedBase Model=LLaVA-NeXT-7B, Confidence Type=Verbalized2026.04 | 56.5 | |
| Platt ScalingBase Model=LLaVA-NeXT-7B, Confidence Type=Verbalized2026.04 | 56.5 | |
| Hist. BinningBase Model=Qwen3-VL-2B, Confidence Type=Sampling2026.04 | 56.4 | |
| Isotonic Regr.Base Model=Qwen3-VL-32B, Confidence Type=Verbalized2026.04 | 56.1 | |
| Isotonic Regr.Base Model=Qwen3-VL-2B, Confidence Type=Verbalized2026.04 | 56 | |
| Isotonic Regr.Base Model=LLaVA-NeXT-7B, Confidence Type=Sampling2026.04 | 56 | |
| HAC-GateBase Model=LLaVA-NeXT-7B, Confidence Type=Sampling2026.04 | 55.9 | |
| HAC-PlattBase Model=Qwen3-VL-2B, Confidence Type=Verbalized2026.04 | 55.7 | |
| UncalibratedBase Model=LLaVA-NeXT-7B, Confidence Type=Sampling2026.04 | 55.6 | |
| Platt ScalingBase Model=LLaVA-NeXT-7B, Confidence Type=Sampling2026.04 | 55.6 | |
| UncalibratedBase Model=Qwen3-VL-2B, Confidence Type=Verbalized2026.04 | 54.9 | |
| Platt ScalingBase Model=Qwen3-VL-2B, Confidence Type=Verbalized2026.04 | 54.9 | |
| UncalibratedBase Model=Qwen3-VL-8B, Confidence Type=Sampling2026.04 | 54.9 | |
| Platt ScalingBase Model=Qwen3-VL-8B, Confidence Type=Sampling2026.04 | 54.9 | |
| Hist. BinningBase Model=InternVL3-2B, Confidence Type=Verbalized2026.04 | 54.7 | |
| Isotonic Regr.Base Model=LLaVA-NeXT-34B, Confidence Type=Verbalized2026.04 | 53.6 | |
| Hist. BinningBase Model=LLaVA-NeXT-7B, Confidence Type=Sampling2026.04 | 53 | |
| UncalibratedBase Model=LLaVA-NeXT-34B, Confidence Type=Verbalized2026.04 | 52.5 | |
| Platt ScalingBase Model=LLaVA-NeXT-34B, Confidence Type=Verbalized2026.04 | 52.5 | |
| Hist. BinningBase Model=LLaVA-NeXT-34B, Confidence Type=Verbalized2026.04 | 52.5 | |
| Isotonic Regr.Base Model=InternVL3-2B, Confidence Type=Verbalized2026.04 | 52.4 | |
| HAC-PlattBase Model=LLaVA-NeXT-7B, Confidence Type=Sampling2026.04 | 52.4 | |
| UncalibratedBase Model=Qwen3-VL-8B, Confidence Type=Verbalized2026.04 | 51.8 | |
| Platt ScalingBase Model=Qwen3-VL-8B, Confidence Type=Verbalized2026.04 | 51.8 | |
| Hist. BinningBase Model=LLaVA-NeXT-34B, Confidence Type=Sampling2026.04 | 51.8 | |
| Isotonic Regr.Base Model=Qwen3-VL-8B, Confidence Type=Sampling2026.04 | 51.6 | |
| Hist. BinningBase Model=Qwen3-VL-2B, Confidence Type=Verbalized2026.04 | 51.1 | |
| Hist. BinningBase Model=InternVL3-8B, Confidence Type=Verbalized2026.04 | 50.9 | |
| Isotonic Regr.Base Model=LLaVA-NeXT-7B, Confidence Type=Verbalized2026.04 | 50.7 | |
| UncalibratedBase Model=Qwen3-VL-32B, Confidence Type=Sampling2026.04 | 50 | |
| Platt ScalingBase Model=Qwen3-VL-32B, Confidence Type=Sampling2026.04 | 50 | |
| Isotonic Regr.Base Model=Qwen3-VL-32B, Confidence Type=Sampling2026.04 | 50 | |
| Hist. BinningBase Model=Qwen3-VL-32B, Confidence Type=Sampling2026.04 | 50 | |
| Isotonic Regr.Base Model=InternVL3-8B, Confidence Type=Verbalized2026.04 | 50 | |
| HAC-PlattBase Model=Qwen3-VL-2B, Confidence Type=Sampling2026.04 | 49.8 | |
| HAC-GateBase Model=Qwen3-VL-2B, Confidence Type=Sampling2026.04 | 49.8 | |
| Hist. BinningBase Model=Qwen3-VL-8B, Confidence Type=Sampling2026.04 | 49.7 | |
| UncalibratedBase Model=InternVL3-8B, Confidence Type=Verbalized2026.04 | 49.1 | |
| Platt ScalingBase Model=InternVL3-8B, Confidence Type=Verbalized2026.04 | 49.1 | |
| HAC-GateBase Model=Qwen3-VL-2B, Confidence Type=Verbalized2026.04 | 48.6 | |
| Isotonic Regr.Base Model=LLaVA-NeXT-34B, Confidence Type=Sampling2026.04 | 48.1 | |
| Hist. BinningBase Model=LLaVA-NeXT-7B, Confidence Type=Verbalized2026.04 | 47.1 | |
| UncalibratedBase Model=InternVL3-38B, Confidence Type=Sampling2026.04 | 46.3 | |
| Platt ScalingBase Model=InternVL3-38B, Confidence Type=Sampling2026.04 | 46.3 | |
| HAC-GateBase Model=InternVL3-38B, Confidence Type=Sampling2026.04 | 46.3 | |
| HAC-PlattBase Model=LLaVA-NeXT-34B, Confidence Type=Sampling2026.04 | 46.1 | |
| HAC-GateBase Model=LLaVA-NeXT-34B, Confidence Type=Verbalized2026.04 | 45.8 | |
| Isotonic Regr.Base Model=InternVL3-38B, Confidence Type=Sampling2026.04 | 44.4 | |
| HAC-PlattBase Model=LLaVA-NeXT-34B, Confidence Type=Verbalized2026.04 | 43.1 | |
| Isotonic Regr.Base Model=Qwen3-VL-8B, Confidence Type=Verbalized2026.04 | 43 | |
| HAC-PlattBase Model=InternVL3-38B, Confidence Type=Sampling2026.04 | 42.3 | |
| HAC-PlattBase Model=Qwen3-VL-32B, Confidence Type=Sampling2026.04 | 42.2 | |
| HAC-GateBase Model=Qwen3-VL-32B, Confidence Type=Sampling2026.04 | 42.2 | |
| Hist. BinningBase Model=Qwen3-VL-32B, Confidence Type=Verbalized2026.04 | 40.2 | |
| Hist. BinningBase Model=Qwen3-VL-8B, Confidence Type=Verbalized2026.04 | 34.2 |