Similarity-as-Evidence: Calibrating Overconfident VLMs for Interpretable and Label-Efficient Medical Active Learning

TL;DR

This paper introduces SaE, a framework that calibrates similarity scores in vision-language models for medical active learning, improving interpretability and reducing overconfidence to enhance sample selection and annotation efficiency.

Contribution

SaE reinterprets similarity scores as evidence using a Dirichlet distribution, enabling better calibration and more effective, interpretable active learning in medical imaging.

Findings

Achieves state-of-the-art accuracy of 82.57% on ten datasets.

Improves calibration with NLL of 0.425 on BTMRI.

Effectively prioritizes rare and ambiguous cases during active learning.

Abstract

Active Learning (AL) reduces annotation costs in medical imaging by selecting only the most informative samples for labeling, but suffers from cold-start when labeled data are scarce. Vision-Language Models (VLMs) address the cold-start problem via zero-shot predictions, yet their temperature-scaled softmax outputs treat text-image similarities as deterministic scores while ignoring inherent uncertainty, leading to overconfidence. This overconfidence misleads sample selection, wasting annotation budgets on uninformative cases. To overcome these limitations, the Similarity-as-Evidence (SaE) framework calibrates text-image similarities by introducing a Similarity Evidence Head (SEH), which reinterprets the similarity vector as evidence and parameterizes a Dirichlet distribution over labels. In contrast to a standard softmax that enforces confident predictions even under weak signals, the Dirichlet formulation explicitly quantifies lack of evidence (vacuity) and conflicting evidence (dissonance), thereby mitigating overconfidence caused by rigid softmax normalization. Building on this, SaE employs a dual-factor acquisition strategy: high-vacuity samples (e.g., rare diseases) are prioritized in early rounds to ensure coverage, while high-dissonance samples (e.g., ambiguous diagnoses) are prioritized later to refine boundaries, providing clinically interpretable selection rationales. Experiments on ten public medical imaging datasets with a 20% label budget show that SaE attains state-of-the-art macro-averaged accuracy of 82.57%. On the representative BTMRI dataset, SaE also achieves superior calibration, with a negative log-likelihood (NLL) of 0.425.