AI Paper Atlas

Research map / Vision-language and multimodal / Visual question answering

Dual-Level Confidence based Implicit Self-Refinement for Medical Visual Question Answering

Meihong Pan, Yefeng Zheng

CVPR 2026 · Vision-language and multimodal › Visual question answering · approach: Attention and prior-robust VQA

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add naive pseudo-labeled test samples as ground truth (ablation)DuCoR baseline (Llama2) · PathVQA closedaccuracy: 87.22 -> 85.41Hurt
add naive pseudo-labels (ablation)DuCoR baseline (Llama2) · PathVQA openrecall: 39.73 -> 40.16No clear effect
add loss-level confidence weighting (ablation)DuCoR (Llama2) · PathVQA closedaccuracy: 85.41 -> 87.95Improved
add feature-level confidence (ablation)DuCoR (Llama2) · PathVQA closedaccuracy: 87.95 -> 89.37Improved
add feature-level confidence (ablation)DuCoR (Llama2) · PathVQAF1: 41.97 -> 41.45Mixed
add prototype contrastive regularization (ablation)DuCoR (Llama2) · PathVQA closedaccuracy: 89.37 -> 89.84Improved
use DuCoR instead of AdaContrast for pseudo-labelingDuCoR · VQA-RADoverall accuracy: 74.70 -> 77.56Improved
use DuCoR adaptation instead of zero-shot transferDuCoR · SLAKE to PathVQArecall: 17.36 -> 24.56Improved
use DuCoR adaptation instead of zero-shot transferDuCoR · PathVQA to SLAKErecall: 28.74 -> 42.32Improved
use DuCoR adaptation instead of zero-shot transferDuCoR · SLAKE to VQA-RADrecall: 45.02 -> 51.81Improved
add DuCoR trainingDuCoR · pseudo-label reliabilityAUROC: 0.6827 -> 0.8575Improved
scale from GPT2 to Llama2 backboneDuCoR · PathVQA closedaccuracy: 88.89 -> 89.84No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy