Faithfulness vs. Safety: Evaluating LLM Behavior Under Counterfactual Medical Evidence

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mo, Kaijie, Venkatayogi, Siddhartha, Shaib, Chantal, Kouzy, Ramez, Xu, Wei, Wallace, Byron C., Li, Junyi Jessy
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914487804624896
author Mo, Kaijie
Venkatayogi, Siddhartha
Shaib, Chantal
Kouzy, Ramez
Xu, Wei
Wallace, Byron C.
Li, Junyi Jessy
author_facet Mo, Kaijie
Venkatayogi, Siddhartha
Shaib, Chantal
Kouzy, Ramez
Xu, Wei
Wallace, Byron C.
Li, Junyi Jessy
contents In high-stakes domains like medicine, it may be generally desirable for models to faithfully adhere to the context provided. But what happens if the context does not align with model priors or safety protocols? In this paper, we investigate how LLMs behave and reason when presented with counterfactual (or even adversarial) medical evidence. We first construct MedCounterFact, a counterfactual medical QA dataset that requires the models to answer clinical comparison questions (i.e., judge the efficacy of certain treatments, with evidence consisting of randomized controlled trials provided as context). In MedCounterFact, real-world medical interventions within the questions and evidence are systematically replaced with four types of counterfactual stimuli, ranging from unknown words to toxic substances. Our evaluation across multiple frontier LLMs on MedCounterFact reveals that in the presence of counterfactual evidence, existing models overwhelmingly accept such "evidence" at face value even when it is dangerous or implausible, and provide confident and uncaveated answers. While it may be prudent to draw a boundary between faithfulness and safety, our findings suggest that models arguably overemphasize the former.
format Preprint
id arxiv_https___arxiv_org_abs_2601_11886
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Faithfulness vs. Safety: Evaluating LLM Behavior Under Counterfactual Medical Evidence
Mo, Kaijie
Venkatayogi, Siddhartha
Shaib, Chantal
Kouzy, Ramez
Xu, Wei
Wallace, Byron C.
Li, Junyi Jessy
Computation and Language
In high-stakes domains like medicine, it may be generally desirable for models to faithfully adhere to the context provided. But what happens if the context does not align with model priors or safety protocols? In this paper, we investigate how LLMs behave and reason when presented with counterfactual (or even adversarial) medical evidence. We first construct MedCounterFact, a counterfactual medical QA dataset that requires the models to answer clinical comparison questions (i.e., judge the efficacy of certain treatments, with evidence consisting of randomized controlled trials provided as context). In MedCounterFact, real-world medical interventions within the questions and evidence are systematically replaced with four types of counterfactual stimuli, ranging from unknown words to toxic substances. Our evaluation across multiple frontier LLMs on MedCounterFact reveals that in the presence of counterfactual evidence, existing models overwhelmingly accept such "evidence" at face value even when it is dangerous or implausible, and provide confident and uncaveated answers. While it may be prudent to draw a boundary between faithfulness and safety, our findings suggest that models arguably overemphasize the former.
title Faithfulness vs. Safety: Evaluating LLM Behavior Under Counterfactual Medical Evidence
topic Computation and Language
url https://arxiv.org/abs/2601.11886