-
The Alignment Auditor: A Bayesian Framework for Verifying and Refining LLM Objectives
Introducing a new framework for auditing the alignment of Large Language Models through Bayesian inference
-
Improving ARDS Diagnosis Through Context-Aware Concept Bottleneck Models
Combining ICU data with LLM-extracted concepts from clinical notes and imaging to improve ARDS diagnosis accuracy
-
Concept-Driven Off-Policy Evaluation
Rethinking OPE using interpretable concepts for low-variance, transparent, and actionable evaluation