The Over-Certainty Phenomenon in Modern Test-Time Adaptation Algorithms

TL;DR

This paper identifies the over-certainty problem in test-time adaptation algorithms, which leads to overconfident predictions under domain shifts, and proposes a certainty regularizer to improve calibration without sacrificing accuracy.

Contribution

It introduces a novel certainty regularizer that dynamically adjusts pseudo-label confidence to mitigate over-certainty in test-time adaptation.

Findings

Achieves state-of-the-art calibration metrics (Expected Calibration Error, Negative Log Likelihood)

Maintains accuracy comparable to existing methods

Addresses the over-certainty phenomenon effectively

Abstract

When neural networks are confronted with unfamiliar data that deviate from their training set, this signifies a domain shift. While these networks output predictions on their inputs, they typically fail to account for their level of familiarity with these novel observations. Prevailing works navigate test-time adaptation with the goal of curtailing model entropy, yet they unintentionally produce models that struggle with sub-optimal calibration-a dilemma we term the over-certainty phenomenon. This over-certainty in predictions can be particularly dangerous in the setting of domain shifts, as it may lead to misplaced trust. In this paper, we propose a solution that not only maintains accuracy but also addresses calibration by mitigating the over-certainty phenomenon. To do this, we introduce a certainty regularizer that dynamically adjusts pseudo-label confidence by accounting for both backbone entropy and logit norm. Our method achieves state-of-the-art performance in terms of Expected Calibration Error and Negative Log Likelihood, all while maintaining parity in accuracy.