Research

I am conducting this dissertation research in the Secure and Assured Intelligent Learning Lab (SAIL Lab) at the University of New Haven under the supervision of Dr. Vahid Behzadan.

My dissertation investigates robustness, safety, and interpretability failures in medical Vision-Language Models (VLMs). The central finding is that models that appear reliable by standard metrics are often exploiting text shortcuts rather than analyzing the medical image. A model can achieve near-perfect consistency while ignoring the chest X-ray entirely.

Interactive failure gallery

I have also published a poster-friendly interactive gallery with representative PSF-Med failure cases. It shows how semantically equivalent clinical questions can trigger contradictory answers on the same chest X-ray while suppressing raw image filenames and internal example IDs.

Open the interactive PSF-Med gallery

Key contributions

• PSF-Med: A benchmark of 82,234 paraphrase pairs across six VLMs showing that rephrasing a clinical question can flip a diagnostic answer up to 54% of the time.
• Consistency-Safety Paradox: Demonstrating that the most consistent models are the most text-reliant, creating a false sense of safety (r = -0.89).
• Mechanistic Diagnosis: Using Sparse Autoencoders to identify a single “formality gate” feature that mediates 67% of flip failures, enabling targeted 0.1%-parameter LoRA repair.
• Deployment Audits: Showing that selective-prediction rules raise accuracy to 97% by selecting text-answerable cases, not image-grounded ones, across three architecture families.

Publications

• PSF-Med: Measuring and Explaining Paraphrase Sensitivity in Medical Vision-Language Models
  B. Sadanandan, V. Behzadan. Under review, 2026.
  arXiv:2602.21428

• Mechanistically Guided LoRA Improves Paraphrase Consistency in Medical Vision-Language Models
  B. Sadanandan, V. Behzadan. CHIL, 2026.
  arXiv:2603.00148

• Consistent but Dangerous: Per-Sample Safety Classification Reveals False Reliability in Medical VLMs
  B. Sadanandan, V. Behzadan. CVPR MedReasoner Workshop, 2026.
  arXiv:2603.20985

• VSF-Med: A Vulnerability Scoring Framework for Medical Vision-Language Models
  B. Sadanandan, V. Behzadan. IEEE ISBI, 2026. Poster.
  arXiv:2507.00052

• When Chain-of-Thought Backfires: Evaluating Prompt Sensitivity in Medical Language Models
  B. Sadanandan, V. Behzadan. 2nd International Conference on Applied Artificial Intelligence (2AI), 2026.
  arXiv:2603.25960

• Predictive Entropy Links Calibration and Paraphrase Sensitivity in Medical VLMs
  B. Sadanandan, V. Behzadan. Under review, 2026.

• Attention Without Grounding: Safety Evaluation of Visual Explanations in Medical VLMs
  B. Sadanandan, V. Behzadan. Under review, 2026.

• Consistency Is Not Safety: Family-Specific Deployment Audits for Medical VLMs
  B. Sadanandan, V. Behzadan. Under review, 2026.

Datasets & Code

• PSF-Med Benchmark: 82K paraphrase pairs, six VLMs, and three chest X-ray datasets: MIMIC-CXR, PadChest, and VinDr-CXR.
• Models: Base, targeted LoRA, and full LoRA checkpoints on Hugging Face.

News

• Apr 2026: PSF-Med poster at SMLM, Yale.
• Apr 2026: Mechanistically Guided LoRA Improves Paraphrase Consistency accepted at CHIL 2026.
• Mar 2026: Chain-of-Thought paper accepted at 2AI 2026.
• Mar 2026: Consistent but Dangerous, CVPR 2026 MedReasoner paper accepted.
• Feb 2026: VSF-Med poster accepted at IEEE ISBI 2026.