AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

ViCoR: Estimating Visual Necessity via Counterfactual Residuals for Multimodal Medical Data Selection

Haotian Gan, Yudong Li, Xintian Li, Jiarui Han, Weidong Tang

NeurIPS 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Visual instruction tuning · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes ViCoR, a data selection method for multimodal medical fine-tuning that scores samples by how visually necessary they are, using counterfactual residuals.

Problem

Medical vision-language models suffer hidden stratification and modality imbalance during fine-tuning, where linguistic priors suppress visually grounded learning and yield weakly grounded rationales.

Approach

Estimates visual necessity of each training sample via counterfactual residuals to select data that enforces visual grounding, rather than scoring samples by generic image-text quality or difficulty (the abstract is truncated before details of the scoring).

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace counterfactual residual with joint-gradient score (ablation)ViCoR · GEMeX-ThinkVG, GMAI-MMBenchaccuracy: 42.94 -> 51.56Improved
add shortcut-dominance weighting (ablation)ViCoR · GEMeX-ThinkVG, GMAI-MMBenchaccuracy: 41.10 -> 51.56Improved
add residual-direction weighting (ablation)ViCoR · GEMeX-ThinkVG, MedBLINKaccuracy: 58.69 -> 63.68Improved
add residual thresholding (ablation)ViCoR · GEMeX-ThinkVG, CheXbench multi-disease accuracyaccuracy: 45.71 -> 46.84Improved
add medical supervision calibration (ablation)ViCoR · GEMeX-ThinkVG, CheXbench multi-disease accuracyaccuracy: 42.90 -> 46.84Improved
add diversity-aware reranking (ablation)ViCoR · GEMeX-ThinkVG, CheXbench multi-disease accuracyaccuracy: 42.25 -> 46.84Improved
rank by residual norm only instead of full ViCoR score (ablation)ViCoR · GEMeX-ThinkVG, CheXbench multi-disease accuracyaccuracy: 44.92 -> 46.84Hurt
calibration plus diversity without counterfactual residual scoring (ablation)ViCoR · GEMeX-ThinkVG, CheXbench multi-disease accuracyaccuracy: 40.82 -> 46.84Hurt
select 20% subset with ViCoR instead of full-data fine-tuningQwen3-VL-4B · CXR-CounterFact-Single cross-benchmark avgaverage accuracy: 61.63 -> 65.29Improved
random 20% subset instead of full dataQwen3-VL-4B · S-Chain cross-benchmark avgaverage accuracy: 62.53 -> 50.40Hurt
ViCoR selection vs best other 20% selection methodInternVL3.5-4B · OmniMedVQA on GEMeX-ThinkVGaccuracy: 88.35 -> 84.18Hurt
replace mean-image shortcut control with blank, shuffled or wrong image (ablation)ViCoR scoring · GEMeX-ThinkVG ranking stabilitySpearman rho: 0.9675 to 0.9686No 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