Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2511.16625 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911554880929792 |
|---|---|
| author | Hossain, Elias Nipu, Md Mehedi Hasan Sheikh, Maleeha Rana, Rajib Neupane, Subash Yousefi, Niloofar |
| author_facet | Hossain, Elias Nipu, Md Mehedi Hasan Sheikh, Maleeha Rana, Rajib Neupane, Subash Yousefi, Niloofar |
| contents | We propose MedBayes-Lite, a lightweight Bayesian enhancement for transformer-based clinical language models that improves reliability through uncertainty-aware prediction. The framework operates without retraining, architectural modification, or additional trainable parameters, and integrates three components: Bayesian Embedding Calibration via Monte Carlo dropout, Uncertainty-Weighted Attention for reliability-aware token aggregation, and Confidence-Guided Decision Shaping for abstention under uncertainty. Across MedQA, PubMedQA, and MIMIC-III, MedBayes-Lite improves calibration and trustworthiness, reducing overconfidence by 32--48\%. In simulated clinical settings, it further supports safer decision-making by flagging uncertain predictions for human review, particularly under distribution shift. For closed API models, the framework remains applicable through sampling-based predictive uncertainty and confidence-guided abstention, while full embedding- and attention-level uncertainty propagation is evaluated on open-weight transformer models. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_16625 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | MedBayes-Lite: Bayesian Uncertainty Quantification for Safe Clinical Decision Support Hossain, Elias Nipu, Md Mehedi Hasan Sheikh, Maleeha Rana, Rajib Neupane, Subash Yousefi, Niloofar Artificial Intelligence We propose MedBayes-Lite, a lightweight Bayesian enhancement for transformer-based clinical language models that improves reliability through uncertainty-aware prediction. The framework operates without retraining, architectural modification, or additional trainable parameters, and integrates three components: Bayesian Embedding Calibration via Monte Carlo dropout, Uncertainty-Weighted Attention for reliability-aware token aggregation, and Confidence-Guided Decision Shaping for abstention under uncertainty. Across MedQA, PubMedQA, and MIMIC-III, MedBayes-Lite improves calibration and trustworthiness, reducing overconfidence by 32--48\%. In simulated clinical settings, it further supports safer decision-making by flagging uncertain predictions for human review, particularly under distribution shift. For closed API models, the framework remains applicable through sampling-based predictive uncertainty and confidence-guided abstention, while full embedding- and attention-level uncertainty propagation is evaluated on open-weight transformer models. |
| title | MedBayes-Lite: Bayesian Uncertainty Quantification for Safe Clinical Decision Support |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2511.16625 |