Medical Visual Question Answering on Pooled Medical VQA Datasets Closed (5-fold CV)
37.1ACEUncalibrated
Evaluation Results
| Method | Links | |
|---|---|---|
| UncalibratedBackbone=LLaVA-NeXT-7B, Confidence Type=Verbalized2026.04 | 37.1 | |
| UncalibratedBackbone=LLaVA-NeXT-34B, Confidence Type=Verbalized2026.04 | 24.1 | |
| UncalibratedBackbone=Qwen3-VL-2B, Confidence Type=Sampling2026.04 | 22.4 | |
| UncalibratedBackbone=Qwen3-VL-8B, Confidence Type=Sampling2026.04 | 21.4 | |
| UncalibratedBackbone=Qwen3-VL-32B, Confidence Type=Sampling2026.04 | 21.2 | |
| UncalibratedBackbone=LLaVA-NeXT-34B, Confidence Type=Sampling2026.04 | 21.2 | |
| UncalibratedBackbone=Qwen3-VL-2B, Confidence Type=Verbalized2026.04 | 19.9 | |
| UncalibratedBackbone=InternVL3-2B, Confidence Type=Sampling2026.04 | 19.4 | |
| UncalibratedBackbone=InternVL3-2B, Confidence Type=Verbalized2026.04 | 19.2 | |
| UncalibratedBackbone=Qwen3-VL-8B, Confidence Type=Verbalized2026.04 | 18 | |
| UncalibratedBackbone=InternVL3-8B, Confidence Type=Verbalized2026.04 | 17.6 | |
| UncalibratedBackbone=Qwen3-VL-32B, Confidence Type=Verbalized2026.04 | 16.6 | |
| UncalibratedBackbone=LLaVA-NeXT-7B, Confidence Type=Sampling2026.04 | 16.5 | |
| UncalibratedBackbone=InternVL3-38B, Confidence Type=Verbalized2026.04 | 16.4 | |
| HAC-GateBackbone=LLaVA-NeXT-7B, Confidence Type=Verbalized2026.04 | 16.3 | |
| UncalibratedBackbone=InternVL3-8B, Confidence Type=Sampling2026.04 | 15.2 | |
| Platt ScalingBackbone=InternVL3-2B, Confidence Type=Sampling2026.04 | 14.5 | |
| HAC-PlattBackbone=LLaVA-NeXT-7B, Confidence Type=Verbalized2026.04 | 14 | |
| HAC-PlattBackbone=InternVL3-2B, Confidence Type=Sampling2026.04 | 13.5 | |
| Hist. BinningBackbone=LLaVA-NeXT-34B, Confidence Type=Sampling2026.04 | 13.3 | |
| Isotonic Regr.Backbone=LLaVA-NeXT-7B, Confidence Type=Sampling2026.04 | 13.1 | |
| HAC-GateBackbone=LLaVA-NeXT-7B, Confidence Type=Sampling2026.04 | 13.1 | |
| Hist. BinningBackbone=LLaVA-NeXT-7B, Confidence Type=Verbalized2026.04 | 12.9 | |
| HAC-GateBackbone=InternVL3-2B, Confidence Type=Sampling2026.04 | 12.8 | |
| Platt ScalingBackbone=LLaVA-NeXT-34B, Confidence Type=Sampling2026.04 | 12.8 | |
| HAC-PlattBackbone=LLaVA-NeXT-34B, Confidence Type=Verbalized2026.04 | 12.8 | |
| Platt ScalingBackbone=LLaVA-NeXT-7B, Confidence Type=Sampling2026.04 | 12.7 | |
| HAC-PlattBackbone=Qwen3-VL-2B, Confidence Type=Verbalized2026.04 | 12.4 | |
| HAC-GateBackbone=InternVL3-2B, Confidence Type=Verbalized2026.04 | 12.4 | |
| HAC-PlattBackbone=LLaVA-NeXT-7B, Confidence Type=Sampling2026.04 | 12.4 | |
| HAC-GateBackbone=LLaVA-NeXT-34B, Confidence Type=Verbalized2026.04 | 12.4 | |
| Platt ScalingBackbone=LLaVA-NeXT-7B, Confidence Type=Verbalized2026.04 | 12.3 | |
| Isotonic Regr.Backbone=LLaVA-NeXT-34B, Confidence Type=Sampling2026.04 | 12.3 | |
| HAC-PlattBackbone=LLaVA-NeXT-34B, Confidence Type=Sampling2026.04 | 12.3 | |
| UncalibratedBackbone=InternVL3-38B, Confidence Type=Sampling2026.04 | 12.2 | |
| HAC-GateBackbone=InternVL3-38B, Confidence Type=Verbalized2026.04 | 12.1 | |
| Isotonic Regr.Backbone=InternVL3-2B, Confidence Type=Sampling2026.04 | 11.9 | |
| HAC-PlattBackbone=InternVL3-38B, Confidence Type=Verbalized2026.04 | 11.9 | |
| HAC-GateBackbone=LLaVA-NeXT-34B, Confidence Type=Sampling2026.04 | 11.9 | |
| Hist. BinningBackbone=InternVL3-2B, Confidence Type=Sampling2026.04 | 11.7 | |
| Platt ScalingBackbone=InternVL3-38B, Confidence Type=Verbalized2026.04 | 11.7 | |
| Platt ScalingBackbone=InternVL3-8B, Confidence Type=Sampling2026.04 | 11.6 | |
| Hist. BinningBackbone=LLaVA-NeXT-7B, Confidence Type=Sampling2026.04 | 11.6 | |
| HAC-GateBackbone=Qwen3-VL-8B, Confidence Type=Verbalized2026.04 | 11.5 | |
| HAC-GateBackbone=Qwen3-VL-2B, Confidence Type=Verbalized2026.04 | 11.4 | |
| HAC-PlattBackbone=InternVL3-8B, Confidence Type=Sampling2026.04 | 11.4 | |
| Hist. BinningBackbone=Qwen3-VL-2B, Confidence Type=Sampling2026.04 | 11.3 | |
| HAC-GateBackbone=InternVL3-8B, Confidence Type=Sampling2026.04 | 11.3 | |
| HAC-GateBackbone=Qwen3-VL-2B, Confidence Type=Sampling2026.04 | 11.2 | |
| Platt ScalingBackbone=Qwen3-VL-32B, Confidence Type=Verbalized2026.04 | 11.1 | |
| Isotonic Regr.Backbone=InternVL3-2B, Confidence Type=Verbalized2026.04 | 11.1 | |
| Isotonic Regr.Backbone=LLaVA-NeXT-7B, Confidence Type=Verbalized2026.04 | 11.1 | |
| HAC-PlattBackbone=Qwen3-VL-2B, Confidence Type=Sampling2026.04 | 11 | |
| Platt ScalingBackbone=InternVL3-2B, Confidence Type=Verbalized2026.04 | 11 | |
| Hist. BinningBackbone=InternVL3-8B, Confidence Type=Sampling2026.04 | 11 | |
| Isotonic Regr.Backbone=InternVL3-38B, Confidence Type=Verbalized2026.04 | 11 | |
| Hist. BinningBackbone=LLaVA-NeXT-34B, Confidence Type=Verbalized2026.04 | 11 | |
| Platt ScalingBackbone=Qwen3-VL-8B, Confidence Type=Verbalized2026.04 | 10.9 | |
| Isotonic Regr.Backbone=LLaVA-NeXT-34B, Confidence Type=Verbalized2026.04 | 10.9 | |
| Platt ScalingBackbone=Qwen3-VL-8B, Confidence Type=Sampling2026.04 | 10.8 | |
| HAC-PlattBackbone=Qwen3-VL-8B, Confidence Type=Verbalized2026.04 | 10.8 | |
| HAC-PlattBackbone=Qwen3-VL-32B, Confidence Type=Verbalized2026.04 | 10.8 | |
| Platt ScalingBackbone=LLaVA-NeXT-34B, Confidence Type=Verbalized2026.04 | 10.7 | |
| HAC-GateBackbone=Qwen3-VL-32B, Confidence Type=Verbalized2026.04 | 10.6 | |
| Platt ScalingBackbone=Qwen3-VL-2B, Confidence Type=Sampling2026.04 | 10.4 | |
| Platt ScalingBackbone=Qwen3-VL-2B, Confidence Type=Verbalized2026.04 | 10.4 | |
| HAC-PlattBackbone=Qwen3-VL-32B, Confidence Type=Sampling2026.04 | 10.4 | |
| HAC-GateBackbone=Qwen3-VL-32B, Confidence Type=Sampling2026.04 | 10.4 | |
| Hist. BinningBackbone=InternVL3-38B, Confidence Type=Verbalized2026.04 | 10.4 | |
| HAC-PlattBackbone=InternVL3-38B, Confidence Type=Sampling2026.04 | 10.3 | |
| Hist. BinningBackbone=Qwen3-VL-8B, Confidence Type=Verbalized2026.04 | 10.2 | |
| Isotonic Regr.Backbone=InternVL3-8B, Confidence Type=Sampling2026.04 | 10.2 | |
| Platt ScalingBackbone=InternVL3-38B, Confidence Type=Sampling2026.04 | 10.2 | |
| HAC-PlattBackbone=Qwen3-VL-8B, Confidence Type=Sampling2026.04 | 10.1 | |
| HAC-GateBackbone=Qwen3-VL-8B, Confidence Type=Sampling2026.04 | 10.1 | |
| HAC-PlattBackbone=InternVL3-2B, Confidence Type=Verbalized2026.04 | 10.1 | |
| HAC-GateBackbone=InternVL3-38B, Confidence Type=Sampling2026.04 | 10.1 | |
| Hist. BinningBackbone=Qwen3-VL-8B, Confidence Type=Sampling2026.04 | 10 | |
| Isotonic Regr.Backbone=Qwen3-VL-32B, Confidence Type=Verbalized2026.04 | 10 | |
| Isotonic Regr.Backbone=Qwen3-VL-32B, Confidence Type=Sampling2026.04 | 9.9 | |
| Hist. BinningBackbone=InternVL3-2B, Confidence Type=Verbalized2026.04 | 9.9 | |
| Isotonic Regr.Backbone=Qwen3-VL-2B, Confidence Type=Sampling2026.04 | 9.8 | |
| Hist. BinningBackbone=Qwen3-VL-32B, Confidence Type=Sampling2026.04 | 9.8 | |
| Hist. BinningBackbone=InternVL3-8B, Confidence Type=Verbalized2026.04 | 9.8 | |
| Hist. BinningBackbone=Qwen3-VL-2B, Confidence Type=Verbalized2026.04 | 9.7 | |
| Hist. BinningBackbone=Qwen3-VL-32B, Confidence Type=Verbalized2026.04 | 9.7 | |
| Isotonic Regr.Backbone=Qwen3-VL-8B, Confidence Type=Sampling2026.04 | 9.6 | |
| Platt ScalingBackbone=Qwen3-VL-32B, Confidence Type=Sampling2026.04 | 9.6 | |
| HAC-PlattBackbone=InternVL3-8B, Confidence Type=Verbalized2026.04 | 9.6 | |
| Isotonic Regr.Backbone=Qwen3-VL-2B, Confidence Type=Verbalized2026.04 | 9.5 | |
| Isotonic Regr.Backbone=Qwen3-VL-8B, Confidence Type=Verbalized2026.04 | 9.5 | |
| HAC-GateBackbone=InternVL3-8B, Confidence Type=Verbalized2026.04 | 9.5 | |
| Isotonic Regr.Backbone=InternVL3-38B, Confidence Type=Sampling2026.04 | 9.5 | |
| Platt ScalingBackbone=InternVL3-8B, Confidence Type=Verbalized2026.04 | 9.4 | |
| Isotonic Regr.Backbone=InternVL3-8B, Confidence Type=Verbalized2026.04 | 9.3 | |
| Hist. BinningBackbone=InternVL3-38B, Confidence Type=Sampling2026.04 | 9.1 |