Marzyeh Ghassemi
Germeshausen Career Development Professor; Associate Professor
Develops machine learning to understand and improve health in ways that are robust, private, and fair.
Marzyeh Ghassemi's lab at MIT works on medical imaging & healthcare AI. 44 papers since 2023, 24 at top venues, mostly NeurIPS and ICML. 12 PhD students and 1 master's student. 42 alumni, most often to Apple and Layer 6 AI. Applications close .
At a glance
- papers, 2023–2026
- 11/ year
- at T1 venues
- 55%
- PhD students
- 12
- MS students
- 1
- alumni listed
- 42
Papers by yearfrom the lab homepage, 2023–2026
- 8 papers2026 · T1 7 · collecting
- 11 papers2025 · T1 8
- 12 papers2024 · T1 6
- 13 papers2023 · T1 3
T1 venuesOtherVenues: NeurIPS, ICML, ICLR, AAAI/ACM AIES, Cell
Research focuspaper titles that use each topic's words
- Medical imaging & healthcare AI14 papers
- fair1 paper
- understand1 paper
- improve0 papers
- private0 papers
Paper list
From the lab homepage.
20268 papers
- Cell (T1)Avoiding common failures in AI for health and medicine (opens in a new tab)
- ICLR (T1)Complementing Self-Consistency with Cross-Model Disagreement for Uncertainty Quantification (opens in a new tab)
- ICML (T1)Escaping the Mode: Multi-Answer Reinforcement Learning in LMs (opens in a new tab)
- NeurIPS (T1)Poison-then-Hide: Finetuning-Activated Backdoor Attack on Pretrained Vision Encoders (opens in a new tab)
- Forty- International Conference on Machine Learning Position Paper TrackPosition: Benchmarks Cannot Establish Deployment Readiness of Clinical AI (opens in a new tab)
- NeurIPS (T1)Rank-Constrained Adaptation for Reliable Real-World Performance (opens in a new tab)
- ICLR (T1)When Style Breaks Safety: Defending LLMs Against Superficial Style Alignment (opens in a new tab)
- ICLR (T1)WRING Out The Bias: A Rotation-Based Alternative To Projection Debiasing (opens in a new tab)
202511 papers
- NeurIPS (T1)Aggregation Hides Out-of-Distribution Generalization Failures from Spurious Correlations (opens in a new tab)
- NeurIPS (T1)An Investigation of Memorization Risk in Healthcare Foundation Models (opens in a new tab)
- Nature Computational ScienceComputational challenges arising in algorithmic fairness and health equity with generative AI (opens in a new tab)
- NeurIPS (T1)KScope: A Framework for Characterizing the Knowledge Status of Language Models (opens in a new tab)
- NeurIPS (T1)Learning the Wrong Lessons: Syntactic-Domain Spurious Correlations in Language Models (opens in a new tab)
- ICML (T1)LEMoN: Label Error Detection using Multimodal Neighbors (opens in a new tab)
- NeurIPS (T1)On Group Sufficiency Under Label Bias (opens in a new tab)
- Scientific ReportsOpportunistic screening of type 2 diabetes with deep metric learning using electronic health records (opens in a new tab)
- ICML (T1)Speak Easy: Eliciting Harmful Jailbreaks from LLMs with Simple Interactions (opens in a new tab)
- The Lancet Digital HealthTackling algorithmic bias and promoting transparency in health datasets: the STANDING Together consensus recommendations (opens in a new tab)
- CVPR (T1)Vision-Language Models Do Not Understand Negation (opens in a new tab)
202412 papers
- ICML (T1)Asymmetry in Low-Rank Adapters of Foundation Models (opens in a new tab)
- NeurIPS (T1)Improving Subgroup Robustness via Data Selection (opens in a new tab)
- Nature Reviews CancerLimiting bias in AI models for improved and equitable cancer care (opens in a new tab)
- NeurIPS (T1)MDAgents: An Adaptive Collaboration of LLMs for Medical Decision Making (opens in a new tab)
- ICML (T1)Measuring Stochastic Data Complexity with Boltzmann Influence Functions (opens in a new tab)
- Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous TechnologiesMental-llm: Leveraging large language models for mental health prediction via online text data (opens in a new tab)
- NEJM AISettling the Score on Algorithmic Discrimination in Health Care (opens in a new tab)
- NeurIPS (T1)Test-Time Debiasing of Vision-Language Embeddings (opens in a new tab)
- Nature MedicineThe limits of fair medical imaging AI in real-world generalization (opens in a new tab)
- bmjTRIPOD+ AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods (opens in a new tab)
- Nature Computational ScienceUsing labels to limit AI misuse in health (opens in a new tab)
- ICLR (T1)Views Can Be Deceiving: Improved SSL Through Feature Space Augmentation (opens in a new tab)
202313 papers
- NeurIPS (T1)Aging with GRACE: Lifelong Model Editing with Discrete Key-Value Adapters (opens in a new tab)
- MIT Schwarzman College of ComputingAlgorithmic Fairness in Chest X-ray Diagnosis: A Case Study (opens in a new tab)
- ICML (T1)Change is Hard: A Closer Look at Subpopulation Shift (opens in a new tab)
- Conference on Health, Inference, and LearningClinical Relevance Score for Guided Trauma Injury Pattern Discovery with Weakly Supervised $ $β$ $-VAE (opens in a new tab)
- The New England journal of medicineConsidering Biased Data as Informative Artifacts in AI-Assisted Health Care (opens in a new tab)
- Machine Learning for HealthcareDeep Metric Learning for the Hemodynamics Inference with Electrocardiogram Signals (opens in a new tab)
- AAAI/ACM AIESEvaluating the Impact of Social Determinants on Health Prediction (opens in a new tab)
- FAccTIn the Name of Fairness: Assessing the Bias in Clinical Record De-identification (opens in a new tab)
- Science AdvancesJudging facts, judging norms: Training machine learning models to judge humans requires a modified approach to labeling data (opens in a new tab)
- The Lancet Digital HealthMachine learning COVID-19 detection from wearables (opens in a new tab)
- Nature Human BehaviourPresentation matters for AI-generated clinical advice (opens in a new tab)
- TMLR (T1)Risk Sensitive Dead-end Identification in Safety-Critical Offline Reinforcement Learning (opens in a new tab)
Where alumni went42 listed, 40% to industry
Industry 10Academia 15
Most often to
- Apple
- Layer 6 AI
- Amazon
Also to
- Harvard University
- MIT
- Stanford University
Apply to MITFall 2027
MIT EECS Computer Science PhD
- Round
- Single round
- Opens
- —
- GRE
- See program page
- English
- TOEFL/IELTS
- Fee
- USD 90
Confirmed on the program page.
Similar labsnearest by research summary
- USMichigan
Jie Liu
- Genomics & computational biology
- USNYU
Romain Lopez
- Genomics & computational biology
- USMIT
Ericmoore Jossou
- Proteins, drugs & materials
- CAToronto
Gary Bader
- Medical imaging & healthcare AI
- USUCLA
Yuzhe Yang
- Medical imaging & healthcare AI
- USUCSD
Tiffany Amariuta
- Medical imaging & healthcare AI
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- The lab homepage
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
MITcloses