This work studies a medical example in which a model is asked to assign resource-allocation probabilities to two people given brief clinical context, and then sees the same scenario with a single extra sentence containing contrasting patient information, showing the context-dependent effect of patient information in a sensitive medical use case.
Abstract
Large language models are being incorporated into sensitive and important decision-making processes across nearly all fields. While prior work studies model bias around inputs and scenario framing, models can also behave in unexpected and undesirable ways due to context accumulated over their deployment. In this work, we study a medical example in which a model is asked to assign resource-allocation probabilities to two people given brief clinical context, and then sees the same scenario with a single extra sentence containing contrasting patient information, either with or without its previous response in context. Across three of four tested models, the paired-context and independent-inference experiments have different probability shifts, often in opposite directions (in favor of Person B vs. in favor of Person A) when new information is provided. We include additional paired-context experiments to show the effect of varying attributes across scenario axes. Our findings show the context-dependent effect of patient information in a sensitive medical use case. More broadly, our work shows the importance of carefully incorporating LLM-based systems into decision-making processes, context engineering, and further model behavioral studies.
Large Language Models (LLMs) have achieved strong performance in medical question answering and clinical reasoning tasks. However, their reliability under uncertainty remains poorly understood which raises critical concerns for deployment in high-stakes clinical settings. In such environments, incorrect predictions are inherently risky, but confident incorrect predictions can be particularly harmful as they may mislead clinical decision-making. In this paper, we conduct a systematic behavioral analysis of LLMs under clinical information uncertainty. We propose an evaluation framework based on the MedMCQA dataset consisting of two complementary uncertainty settings. First, we introduce linguistic uncertainty cues through prompt modifications to simulate ambiguous clinical contexts. Second, we construct an answer removal setting, wherein the correct option is deliberately excluded mandating the model to recognize insufficient information and abstain. We analyze both model accuracy and confidence behavior using multiple calibration metrics including calibration gap, Expected Calibration Error (ECE), and Unsafe Confident Error Rate (UCER) across 500 medical questions. Our results reveal a consistent failure mode, i.e., although accuracy degrades under increasing uncertainty, model confidence remains misaligned with accuracy. This leads to a substantial increase in unsafe confident errors, indicating that model confidence remains largely insensitive to clinically meaningful information loss. Furthermore, we observe significant variation across models in their ability to abstain when the correct answer is unavailable, with some models persistently producing high confidence hallucinated answers. These findings expose critical limitations in the epistemic reliability of current LLMs and highlight the need for uncertainty aware evaluation methods prior to their deployment in clinical workflows.
Maryam Tahermazandarani, Adnan Mahmood, Fahmida Islam et al.· 0 citations
Large language models (LLMs) exhibit strong reasoning and world-knowledge capabilities, yet often struggle to gather information effectively across the multi-turn interactions required in sequential decision-making settings. We introduce Amortised Sequential Information Gathering (ASIG), a fine-tuning approach that amortises Bayesian Experimental Design (BED) into LLM policies via a multi-turn extension of Group Relative Policy Optimisation with an Expected Information Gain reward. Evaluated on the 20 Questions task, ASIG more than doubles the success rate of the 7B base model and reduces inference cost by over $25\times$ relative to BED-LLM, a competitive inference-time baseline. Applied to MediQ, a medical diagnosis benchmark unseen during training, ASIG improves information-seeking performance at the 7B scale, suggesting that the learned strategies can transfer out of distribution. Our findings show that amortising BED into LLM policies provides an effective and computationally efficient approach to sequential information gathering.
Jakob Hartmann, James Harvey, Jhonathan Navott et al.· 0 citations
Medical multiple-choice question answering (MCQA) benchmarks implicitly assume that large language models (LLMs) should always commit to an answer. However, in clinical practice, uncertainty is pervasive and abstaining is often the safest action. We introduce MedQAbstain , a benchmark explicitly designed to evaluate medical abstention under uncertainty. MedQAbstain repurposes standard medical MCQA datasets by removing the gold answer and introducing an explicit “I ab-stain” option, framed as a safety-critical decision with clinical consequences. The benchmark supports systematic analysis across ab-stention regimes, distractor complexity, and input modalities, and elicits self-reported model confidence to study calibration. Across all settings, we find that state-of-the-art LLMs systematically overcommit, rarely abstaining even when the question itself is hidden. These results reveal a fundamental mismatch between LLM behavior and clinical norms, highlighting ab-stention as a critical but overlooked dimension of medical decision-making evaluation. 1
Alessio Cocchieri, Luca Ragazzi, Giuseppe Tagliavini et al.· Annual Meeting of the Associ...· 2 citations
Neural networks increasingly combine data across populations, time periods, and operating conditions to improve generalization. This raises a reliability question: whether a model refitted on pooled data preserves an action ordering supported by both sources. Case-Based Decision Theory (CBDT) formalizes this requirement through its composition axiom, which requires source-supported preferences to survive their union. We study when this property holds for fixed-representation neural networks with ordinary least squares (OLS) output heads. First, we show that pooled refitting recomputes the inverse-Gram geometry used to weight source evidence, which can reverse shared preferences, and derive exact and approximate preservation conditions. Next, we introduce a scale-invariant Gram mismatch measure for prioritizing candidate pools and geometry-oriented regularization for shaping source geometry during training. Finally, we develop a three-stage audit that traces strict pairwise reversals through decision changes to task-defined utility loss. Experiments spanning a load-based bidding proxy and medical and financial decision proxies reveal stable and reversal-prone pooling regimes: the load audit identifies a measurable nonzero class of source-consensus-relative harmful decisions under the proxy utility, while cross-domain audits show that comparable mismatch can correspond to sharply different preservation rates. Geometry-oriented objectives occupy distinct descriptive accuracy-consistency-geometry-harm operating points. Together, the framework makes compositional reliability measurable and operational through screening, analytic certification, geometry-oriented training, and decision-consequence auditing.
Yanli Yan, Yuanzheng Li, Yong Zhao et al.· 0 citations
As large language models (LLMs) grow more capable, they are increasingly deployed in context-rich settings where task inputs are often accompanied by long, partially irrelevant context. In a controlled setting, we find that state-of-the-art models often appear robust to task-irrelevant context at the aggregate level: prepending it to benchmark questions causes little change in overall accuracy. This aggregate stability, however, masks significant per-example instability. Even semantically meaningless pseudo-words, formed by randomly combining characters, can markedly shift model predictions on a small fraction of examples, degrading performance on some while improving it on others. This two-sided effect holds consistently across a wide range of models and datasets, yet the affected examples are largely model-specific. We further show that this instability is modulated by context type, context length, test-time compute, and model development stage. Together, our findings reveal context-induced tail risks concealed by aggregate accuracy, motivating per-example reliability evaluation of language models.
Large language models (LLMs) demonstrate expert-level performance in various medical scenarios, yet their outputs can exhibit bias against groups or individuals with specific sensitive attributes, posing risks to patient safety and undermining trust in LLMs for healthcare. Recent research suggests that prompt engineering offers a convenient way to adjust model outputs, with the potential to mitigate such biases. However, there is a lack of empirical studies that systematically examine the effectiveness of prompt engineering and its trade-offs among fairness, accuracy and inference overhead. To fill this gap, we empirically evaluate five widely used prompting strategies across five influential LLMs in the latest medical bias benchmark. Results reveal substantial heterogeneity in both effectiveness and overhead across models, with no strategy proving universally effective and some even exacerbating bias. Chain-of-thought prompting yields the largest reduction, lowering the average gap across all scenarios by 2.4 percentage points, where the largest reduction is 6.2 percentage points, obtained in the DeepSeek-V3.1-sex case. Furthermore, the results of the McNemar test also show that it achieves the largest number of significant bias reduction cases (8/15), primarily by improving performance on unprivileged groups. These findings provide practical guidance for the fair deployment of LLMs in healthcare and highlight that mitigating medical bias remains a challenging problem requiring sustained efforts from both the artificial intelligence (AI) and medical communities. To support future research on fair AI in healthcare, we shall release all results and source code. This article is part of the theme issue 'Safe, secure and robust AI for safety-critical systems'.
Ying Xiao, Zhenpeng Chen, Jie M. Zhang· Philosophical transactions....· 1 citation
Known for his clear and elegant writing style, Bertsekas shaped fields from control and optimization to large-scale computation and artificial intelligence.
MIT News · Artificial Intelligence· news.mit.eduJul 7, 2026
The professor of physics and inaugural director of the NSF AI Institute for Artificial Intelligence and Fundamental Interactions will lead LNS and continue his research in particle physics.
