Medical Visual Question Answering on SLAKE Open (ACE)
45.2ACELLaVA-NeXT-34B
Evaluation Results
| Method | Links | |
|---|---|---|
| LLaVA-NeXT-34BCalibration=Uncalibrated, Confidence=Verbalized2026.04 | 45.2 | |
| LLaVA-NeXT-7BCalibration=Uncalibrated, Confidence=Verbalized2026.04 | 44.6 | |
| InternVL3-38BCalibration=Uncalibrated, Confidence=Verbalized2026.04 | 35.1 | |
| InternVL3-2BCalibration=Uncalibrated, Confidence=Verbalized2026.04 | 33.9 | |
| Qwen3-VL-2BCalibration=Uncalibrated, Confidence=Verbalized2026.04 | 33.7 | |
| Qwen3-VL-8BCalibration=Uncalibrated, Confidence=Sampling2026.04 | 33.1 | |
| Qwen3-VL-8BCalibration=Uncalibrated, Confidence=Verbalized2026.04 | 29.9 | |
| LLaVA-NeXT-34BCalibration=Uncalibrated, Confidence=Sampling2026.04 | 29.9 | |
| Qwen3-VL-2BCalibration=Uncalibrated, Confidence=Sampling2026.04 | 28 | |
| Qwen3-VL-32BCalibration=Uncalibrated, Confidence=Sampling2026.04 | 26.5 | |
| InternVL3-8BCalibration=Uncalibrated, Confidence=Verbalized2026.04 | 26.5 | |
| Qwen3-VL-32BCalibration=Uncalibrated, Confidence=Verbalized2026.04 | 25.9 | |
| LLaVA-NeXT-7BCalibration=Uncalibrated, Confidence=Sampling2026.04 | 25 | |
| LLaVA-NeXT-34BCalibration=Hist. Binning, Confidence=Sampling2026.04 | 23.9 | |
| InternVL3-2BCalibration=Uncalibrated, Confidence=Sampling2026.04 | 23.2 | |
| InternVL3-38BCalibration=Uncalibrated, Confidence=Sampling2026.04 | 22.8 | |
| LLaVA-NeXT-34BCalibration=Isotonic Regr., Confidence=Sampling2026.04 | 22.7 | |
| LLaVA-NeXT-34BCalibration=Platt Scaling, Confidence=Sampling2026.04 | 22.3 | |
| InternVL3-8BCalibration=Uncalibrated, Confidence=Sampling2026.04 | 22 | |
| LLaVA-NeXT-34BCalibration=HAC-Gate, Confidence=Sampling2026.04 | 20 | |
| LLaVA-NeXT-34BCalibration=Isotonic Regr., Confidence=Verbalized2026.04 | 19.8 | |
| LLaVA-NeXT-34BCalibration=HAC-Platt, Confidence=Sampling2026.04 | 19.7 | |
| LLaVA-NeXT-34BCalibration=Platt Scaling, Confidence=Verbalized2026.04 | 19.6 | |
| LLaVA-NeXT-34BCalibration=Hist. Binning, Confidence=Verbalized2026.04 | 19 | |
| LLaVA-NeXT-34BCalibration=HAC-Gate, Confidence=Verbalized2026.04 | 18 | |
| LLaVA-NeXT-34BCalibration=HAC-Platt, Confidence=Verbalized2026.04 | 16.6 | |
| Qwen3-VL-2BCalibration=HAC-Gate, Confidence=Sampling2026.04 | 14.4 | |
| Qwen3-VL-2BCalibration=HAC-Platt, Confidence=Verbalized2026.04 | 14.3 | |
| Qwen3-VL-2BCalibration=HAC-Platt, Confidence=Sampling2026.04 | 14 | |
| LLaVA-NeXT-7BCalibration=Platt Scaling, Confidence=Verbalized2026.04 | 14 | |
| InternVL3-2BCalibration=HAC-Gate, Confidence=Verbalized2026.04 | 13.9 | |
| InternVL3-38BCalibration=HAC-Gate, Confidence=Verbalized2026.04 | 13.9 | |
| Qwen3-VL-2BCalibration=HAC-Gate, Confidence=Verbalized2026.04 | 13.8 | |
| Qwen3-VL-32BCalibration=Hist. Binning, Confidence=Verbalized2026.04 | 13.7 | |
| Qwen3-VL-32BCalibration=Platt Scaling, Confidence=Sampling2026.04 | 13.5 | |
| Qwen3-VL-8BCalibration=Hist. Binning, Confidence=Sampling2026.04 | 13.2 | |
| Qwen3-VL-32BCalibration=Hist. Binning, Confidence=Sampling2026.04 | 13.2 | |
| Qwen3-VL-32BCalibration=HAC-Platt, Confidence=Sampling2026.04 | 13.2 | |
| InternVL3-38BCalibration=Isotonic Regr., Confidence=Sampling2026.04 | 13 | |
| Qwen3-VL-2BCalibration=Hist. Binning, Confidence=Verbalized2026.04 | 12.8 | |
| Qwen3-VL-8BCalibration=Platt Scaling, Confidence=Sampling2026.04 | 12.8 | |
| Qwen3-VL-8BCalibration=Platt Scaling, Confidence=Verbalized2026.04 | 12.8 | |
| InternVL3-38BCalibration=HAC-Platt, Confidence=Verbalized2026.04 | 12.7 | |
| LLaVA-NeXT-7BCalibration=Isotonic Regr., Confidence=Sampling2026.04 | 12.7 | |
| Qwen3-VL-2BCalibration=Platt Scaling, Confidence=Sampling2026.04 | 12.6 | |
| InternVL3-2BCalibration=HAC-Platt, Confidence=Verbalized2026.04 | 12.6 | |
| Qwen3-VL-8BCalibration=Isotonic Regr., Confidence=Sampling2026.04 | 12.5 | |
| InternVL3-8BCalibration=Isotonic Regr., Confidence=Sampling2026.04 | 12.5 | |
| InternVL3-38BCalibration=Platt Scaling, Confidence=Sampling2026.04 | 12.5 | |
| InternVL3-2BCalibration=Platt Scaling, Confidence=Verbalized2026.04 | 12.4 | |
| InternVL3-8BCalibration=Hist. Binning, Confidence=Sampling2026.04 | 12.4 | |
| InternVL3-38BCalibration=Hist. Binning, Confidence=Verbalized2026.04 | 12.4 | |
| Qwen3-VL-8BCalibration=HAC-Gate, Confidence=Verbalized2026.04 | 12.3 | |
| InternVL3-8BCalibration=Platt Scaling, Confidence=Verbalized2026.04 | 12.3 | |
| Qwen3-VL-32BCalibration=HAC-Gate, Confidence=Sampling2026.04 | 12.1 | |
| LLaVA-NeXT-7BCalibration=Isotonic Regr., Confidence=Verbalized2026.04 | 12.1 | |
| Qwen3-VL-8BCalibration=Isotonic Regr., Confidence=Verbalized2026.04 | 12 | |
| InternVL3-38BCalibration=HAC-Platt, Confidence=Sampling2026.04 | 12 | |
| Qwen3-VL-2BCalibration=Hist. Binning, Confidence=Sampling2026.04 | 11.9 | |
| InternVL3-2BCalibration=HAC-Platt, Confidence=Sampling2026.04 | 11.9 | |
| LLaVA-NeXT-7BCalibration=HAC-Platt, Confidence=Verbalized2026.04 | 11.8 | |
| Qwen3-VL-2BCalibration=Isotonic Regr., Confidence=Sampling2026.04 | 11.7 | |
| Qwen3-VL-2BCalibration=Isotonic Regr., Confidence=Verbalized2026.04 | 11.7 | |
| Qwen3-VL-32BCalibration=Isotonic Regr., Confidence=Sampling2026.04 | 11.7 | |
| InternVL3-2BCalibration=Hist. Binning, Confidence=Sampling2026.04 | 11.7 | |
| InternVL3-2BCalibration=Hist. Binning, Confidence=Verbalized2026.04 | 11.7 | |
| InternVL3-2BCalibration=HAC-Gate, Confidence=Sampling2026.04 | 11.7 | |
| InternVL3-8BCalibration=Platt Scaling, Confidence=Sampling2026.04 | 11.7 | |
| InternVL3-38BCalibration=Platt Scaling, Confidence=Verbalized2026.04 | 11.7 | |
| InternVL3-38BCalibration=Hist. Binning, Confidence=Sampling2026.04 | 11.7 | |
| LLaVA-NeXT-7BCalibration=HAC-Gate, Confidence=Verbalized2026.04 | 11.7 | |
| Qwen3-VL-2BCalibration=Platt Scaling, Confidence=Verbalized2026.04 | 11.6 | |
| InternVL3-2BCalibration=Platt Scaling, Confidence=Sampling2026.04 | 11.6 | |
| InternVL3-38BCalibration=Isotonic Regr., Confidence=Verbalized2026.04 | 11.6 | |
| LLaVA-NeXT-7BCalibration=Hist. Binning, Confidence=Sampling2026.04 | 11.6 | |
| LLaVA-NeXT-7BCalibration=HAC-Gate, Confidence=Sampling2026.04 | 11.6 | |
| InternVL3-8BCalibration=Hist. Binning, Confidence=Verbalized2026.04 | 11.5 | |
| LLaVA-NeXT-7BCalibration=HAC-Platt, Confidence=Sampling2026.04 | 11.5 | |
| Qwen3-VL-8BCalibration=HAC-Gate, Confidence=Sampling2026.04 | 11.4 | |
| Qwen3-VL-32BCalibration=Platt Scaling, Confidence=Verbalized2026.04 | 11.4 | |
| InternVL3-2BCalibration=Isotonic Regr., Confidence=Sampling2026.04 | 11.4 | |
| InternVL3-8BCalibration=Isotonic Regr., Confidence=Verbalized2026.04 | 11.4 | |
| InternVL3-2BCalibration=Isotonic Regr., Confidence=Verbalized2026.04 | 11.3 | |
| InternVL3-8BCalibration=HAC-Gate, Confidence=Sampling2026.04 | 11.3 | |
| LLaVA-NeXT-7BCalibration=Hist. Binning, Confidence=Verbalized2026.04 | 11.2 | |
| Qwen3-VL-8BCalibration=HAC-Platt, Confidence=Sampling2026.04 | 11.1 | |
| Qwen3-VL-8BCalibration=HAC-Platt, Confidence=Verbalized2026.04 | 11.1 | |
| LLaVA-NeXT-7BCalibration=Platt Scaling, Confidence=Sampling2026.04 | 11.1 | |
| InternVL3-8BCalibration=HAC-Platt, Confidence=Sampling2026.04 | 11 | |
| InternVL3-38BCalibration=HAC-Gate, Confidence=Sampling2026.04 | 11 | |
| Qwen3-VL-32BCalibration=HAC-Gate, Confidence=Verbalized2026.04 | 10.9 | |
| InternVL3-8BCalibration=HAC-Gate, Confidence=Verbalized2026.04 | 10.5 | |
| Qwen3-VL-8BCalibration=Hist. Binning, Confidence=Verbalized2026.04 | 10.4 | |
| Qwen3-VL-32BCalibration=HAC-Platt, Confidence=Verbalized2026.04 | 10.4 | |
| Qwen3-VL-32BCalibration=Isotonic Regr., Confidence=Verbalized2026.04 | 10 | |
| InternVL3-8BCalibration=HAC-Platt, Confidence=Verbalized2026.04 | 10 |