When Reasoning Hurts Confidence: Demographic Calibration of Vision-Language Models in Medicine

S Xu, R Daneshjou

AAAI/ACM Conference on AI, Ethics, and Society (AIES), 2026

Visual Concept Ranking Uncovers Medical Shortcuts Used by Large Multimodal Models

JD Janizek, S Xu, J Lateef, R Daneshjou

Machine Learning for Healthcare (MLHC), 2026

Paper

Demographic Calibration of Vision-Language Models for Dermatology

S Xu, R Daneshjou

AISTATS Calibration Workshop, 2026

Paper Workshop

BiasICL: In-Context Learning and Demographic Biases of Vision Language Models

S Xu, JD Janizek, Y Jiang, R Daneshjou

MICCAI, 2025

Paper Code

Ethical Obligations to Inform Patients About AI Tool Use

MM Mello, D Char, SH Xu

JAMA, 2025

Paper

SycEval: Evaluating LLM Sycophancy

A Fanous, J Goldberg, A Agarwal, J Lin, A Zhou, S Xu, V Bikia, et al.

AAAI/ACM Conference on AI, Ethics, and Society, 2025

Paper

Automating Pharmacogenomic Annotation: Leveraging Schema Constrained GPT-4 for Variant-Drug Data Extraction

A Fanous, Z Ansari, S Xu, C Thorn, R Daneshjou, T Klein

Genetics in Medicine Open, 2025

Paper

A Framework for Evaluating the Efficacy of Foundation Embedding Models in Healthcare

S Xu, H Gui, V Rotemberg, T Wang, YT Chen, R Daneshjou

medRxiv, 2024

Paper

Directing Generalist Vision-Language Models to Interpret Medical Images Across Populations

LW Sagers, A Shah, S Xu, R Daneshjou, AK Manrai

NeurIPS GenAI for Health Workshop, 2024

Paper

DREAM: A Framework for Discovering Mechanisms Underlying AI Prediction of Protected Attributes

S Gadgil*, AJ DeGrave, JD Janizek, S Xu, L Nwandu, F Fonjungo, S Lee†, R Daneshjou†

medRxiv, 2024 · Earlier version: CVPR DCAMI Workshop, 2024 (Oral, Best Paper Runner-Up)

Paper Code

Behaving Better, Thinking Worse: Sycophancy Across Post-Training Stages

S Xu, K Singh, S Jafry, R Daneshjou, S Koyejo

Under review (NeurIPS)

Benchmarking Multimodal Large Language Models Against Human Consensus for Dermatologic Misinformation Assessment

S Xu, A Fanous, K Singh, A Agarwal, M Le, et al., J Ko, J Lipoff, R Daneshjou

Under review

Is Fitzpatrick-Stratified Accuracy a Valid Fairness Metric? An Item-Response-Theory Audit of Dermatology AI

S Xu, R Daneshjou

Under review (ECCV Workshop)

Projection-Preserving Model Merging for Cross-Domain Medical Imaging

S Xu, R Daneshjou

Under review

Med-REDUCE: Representation Transfer and Efficiency Under Resolution Constraints

V Bikia*, S Xu*, A Skrika, R Park, A Fanous, R Daneshjou

Under review · *equal contribution