Aligning Probabilistic Beliefs under Informative Missingness: LLM Steerability in Clinical Reasoning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kobayashi, Yuta, Jeanselme, Vincent, Joshi, Shalmali
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911523969957888
author Kobayashi, Yuta
Jeanselme, Vincent
Joshi, Shalmali
author_facet Kobayashi, Yuta
Jeanselme, Vincent
Joshi, Shalmali
contents Large Language Models (LLMs) are increasingly deployed for clinical reasoning tasks, which inherently require eliciting calibrated probabilistic beliefs based on available evidence. However, real-world clinical data are frequently incomplete, with missingness patterns often informative of patient prognosis; for example, ordering a rare laboratory test reflects a clinician's latent suspicion. In this work, we investigate whether LLMs can be steered to leverage this informative missingness for prognostic inference. To evaluate how well LLMs align their verbalized probabilistic beliefs with an underlying target distribution, we analyze three common prompt-based interventions: explicit serialization, instruction steering, and in-context learning. We introduce a bias-variance decomposition of the log-loss to clarify the mechanisms driving gains in predictive performance. Using a real-world intensive care testbed, we find that while explicit structural steering and in-context learning can improve probabilistic alignment, the models do not natively leverage informative missingness without careful interventions.
format Preprint
id arxiv_https___arxiv_org_abs_2512_00479
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Aligning Probabilistic Beliefs under Informative Missingness: LLM Steerability in Clinical Reasoning
Kobayashi, Yuta
Jeanselme, Vincent
Joshi, Shalmali
Artificial Intelligence
Large Language Models (LLMs) are increasingly deployed for clinical reasoning tasks, which inherently require eliciting calibrated probabilistic beliefs based on available evidence. However, real-world clinical data are frequently incomplete, with missingness patterns often informative of patient prognosis; for example, ordering a rare laboratory test reflects a clinician's latent suspicion. In this work, we investigate whether LLMs can be steered to leverage this informative missingness for prognostic inference. To evaluate how well LLMs align their verbalized probabilistic beliefs with an underlying target distribution, we analyze three common prompt-based interventions: explicit serialization, instruction steering, and in-context learning. We introduce a bias-variance decomposition of the log-loss to clarify the mechanisms driving gains in predictive performance. Using a real-world intensive care testbed, we find that while explicit structural steering and in-context learning can improve probabilistic alignment, the models do not natively leverage informative missingness without careful interventions.
title Aligning Probabilistic Beliefs under Informative Missingness: LLM Steerability in Clinical Reasoning
topic Artificial Intelligence
url https://arxiv.org/abs/2512.00479