Medical Visual Question Answering on Pooled Medical VQA Closed-ended
0.705AUROCHAC-Gate
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| HAC-GateBackbone=InternVL3-2B, Confidence Source=Sampling2026.04 | 0.705 | — | — | |
| HAC-PlattBackbone=InternVL3-2B, Confidence Source=Sampling2026.04 | 0.702 | — | — | |
| HAC-GateBackbone=InternVL3-8B, Confidence Source=Sampling2026.04 | 0.678 | — | — | |
| HAC-PlattBackbone=InternVL3-8B, Confidence Source=Sampling2026.04 | 0.677 | — | — | |
| HAC-PlattBackbone=InternVL3-38B, Confidence Source=Sampling2026.04 | 0.661 | — | — | |
| HAC-GateBackbone=InternVL3-38B, Confidence Source=Sampling2026.04 | 0.66 | — | — | |
| HAC-GateBackbone=InternVL3-2B, Confidence Source=Verbalized2026.04 | 0.658 | — | — | |
| HAC-PlattBackbone=InternVL3-38B, Confidence Source=Verbalized2026.04 | 0.657 | — | — | |
| HAC-GateBackbone=InternVL3-38B, Confidence Source=Verbalized2026.04 | 0.657 | — | — | |
| HAC-GateBackbone=Qwen3-VL-2B, Confidence Source=Sampling2026.04 | 0.654 | — | — | |
| HAC-PlattBackbone=Qwen3-VL-2B, Confidence Source=Sampling2026.04 | 0.653 | — | — | |
| HAC-PlattBackbone=Qwen3-VL-32B, Confidence Source=Verbalized2026.04 | 0.651 | — | — | |
| HAC-GateBackbone=Qwen3-VL-32B, Confidence Source=Verbalized2026.04 | 0.651 | — | — | |
| HAC-PlattBackbone=InternVL3-2B, Confidence Source=Verbalized2026.04 | 0.649 | — | — | |
| HAC-GateBackbone=Qwen3-VL-2B, Confidence Source=Verbalized2026.04 | 0.648 | — | — | |
| HAC-PlattBackbone=Qwen3-VL-2B, Confidence Source=Verbalized2026.04 | 0.644 | — | — | |
| HAC-PlattBackbone=Qwen3-VL-8B, Confidence Source=Sampling2026.04 | 0.639 | — | — | |
| HAC-GateBackbone=Qwen3-VL-8B, Confidence Source=Sampling2026.04 | 0.639 | — | — | |
| HAC-PlattBackbone=LLaVA-NeXT-34B, Confidence Source=Sampling2026.04 | 0.62 | — | — | |
| HAC-GateBackbone=LLaVA-NeXT-34B, Confidence Source=Sampling2026.04 | 0.619 | — | — | |
| HAC-PlattBackbone=Qwen3-VL-32B, Confidence Source=Sampling2026.04 | 0.61 | — | — | |
| HAC-GateBackbone=Qwen3-VL-32B, Confidence Source=Sampling2026.04 | 0.61 | — | — | |
| HAC-PlattBackbone=Qwen3-VL-8B, Confidence Source=Verbalized2026.04 | 0.608 | — | — | |
| HAC-GateBackbone=Qwen3-VL-8B, Confidence Source=Verbalized2026.04 | 0.608 | — | — | |
| HAC-GateBackbone=LLaVA-NeXT-7B, Confidence Source=Sampling2026.04 | 0.598 | — | — | |
| HAC-PlattBackbone=LLaVA-NeXT-7B, Confidence Source=Sampling2026.04 | 0.597 | — | — | |
| UncalibratedBackbone=Qwen3-VL-2B, Confidence Source=Sampling2026.04 | 0.586 | — | — | |
| Platt ScalingBackbone=Qwen3-VL-2B, Confidence Source=Sampling2026.04 | 0.586 | — | — | |
| Hist. BinningBackbone=Qwen3-VL-2B, Confidence Source=Sampling2026.04 | 0.583 | — | — | |
| Isotonic Regr.Backbone=Qwen3-VL-2B, Confidence Source=Sampling2026.04 | 0.577 | — | — | |
| HAC-PlattBackbone=LLaVA-NeXT-34B, Confidence Source=Verbalized2026.04 | 0.574 | — | — | |
| HAC-GateBackbone=LLaVA-NeXT-34B, Confidence Source=Verbalized2026.04 | 0.574 | — | — | |
| UncalibratedBackbone=Qwen3-VL-2B, Confidence Source=Verbalized2026.04 | 0.559 | — | — | |
| Platt ScalingBackbone=Qwen3-VL-2B, Confidence Source=Verbalized2026.04 | 0.559 | — | — | |
| Isotonic Regr.Backbone=Qwen3-VL-2B, Confidence Source=Verbalized2026.04 | 0.557 | — | — | |
| HAC-GateBackbone=LLaVA-NeXT-7B, Confidence Source=Verbalized2026.04 | 0.556 | — | — | |
| HAC-PlattBackbone=LLaVA-NeXT-7B, Confidence Source=Verbalized2026.04 | 0.547 | — | — | |
| HAC-PlattBackbone=InternVL3-8B, Confidence Source=Verbalized2026.04 | 0.545 | — | — | |
| HAC-GateBackbone=InternVL3-8B, Confidence Source=Verbalized2026.04 | 0.545 | — | — | |
| Hist. BinningBackbone=Qwen3-VL-2B, Confidence Source=Verbalized2026.04 | 0.503 | — | — | |
| HAC-GateModel=Qwen3-VL-8B, Calibration Method=HAC-Gate2026.04 | — | 0.639 | 0.608 | |
| HAC-GateModel=InternVL3-8B, Calibration Method=HAC-Gate2026.04 | — | 0.678 | 0.545 | |
| HAC-GateModel=LLaVA-NeXT-7B, Calibration Method=HAC-Gate2026.04 | — | 0.598 | 0.556 | |
| HAC-PlattModel=Qwen3-VL-8B, Calibration Method=HAC-Platt2026.04 | — | 0.639 | 0.608 | |
| HAC-PlattModel=InternVL3-8B, Calibration Method=HAC-Platt2026.04 | — | 0.677 | 0.545 | |
| HAC-PlattModel=LLaVA-NeXT-7B, Calibration Method=HAC-Platt2026.04 | — | 0.597 | 0.547 | |
| Isotonic Regr.Model=Qwen3-VL-8B, Calibration Method=Isotonic Regr.2026.04 | — | 0.556 | 0.604 | |
| Isotonic Regr.Model=InternVL3-8B, Calibration Method=Isotonic Regr.2026.04 | — | 0.66 | 0.522 | |
| Isotonic Regr.Model=LLaVA-NeXT-7B, Calibration Method=Isotonic Regr.2026.04 | — | 0.587 | 0.562 | |
| Platt ScalingModel=Qwen3-VL-8B, Calibration Method=Platt Scaling2026.04 | — | 0.551 | 0.6 | |
| Platt ScalingModel=InternVL3-8B, Calibration Method=Platt Scaling2026.04 | — | 0.663 | 0.522 | |
| Platt ScalingModel=LLaVA-NeXT-7B, Calibration Method=Platt Scaling2026.04 | — | 0.598 | 0.558 | |
| UncalibratedModel=Qwen3-VL-8B, Calibration Method=Uncalibrated2026.04 | — | 0.551 | 0.6 | |
| UncalibratedModel=InternVL3-8B, Calibration Method=Uncalibrated2026.04 | — | 0.663 | 0.522 | |
| UncalibratedModel=LLaVA-NeXT-7B, Calibration Method=Uncalibrated2026.04 | — | 0.598 | 0.558 |