MIRA: Mutual Information guided calibration for Reliable Test-Time Adaptation
MIRA is a source-free online adaptation method using lightweight Gaussian perturbations and mutual information to improve test-time adaptation under distribution shift.
Standard test-time adaptation relies on entropy minimization which can reinforce overconfident errors. MIRA reuses a feature extraction pass and samples perturbed classifiers around the source model for each target batch. It selects perturbation scale using mutual information and prediction agreement to identify reliable predictions before weighted entropy minimization.
- No source data or retraining required
- Uses last-layer Gaussian perturbations
- Mutual information selects batch-specific scale
- Improves accuracy and calibration on distribution-shift benchmarks
From the original article
1 min read at arxiv.orgTest-time adaptation (TTA) updates a deployed model using only unlabeled target data, but standard entropy-based methods often rely on deterministic confidence and can reinforce overconfident errors under distribution shift.
By Hyeongyu Kim · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.