Bayes-inspired Integration of Pretrained Priors and Few-Shot Evidence for Few-Shot Classification

Few-shot classification aims to adapt a pretrained model to novel classes with limited examples. While current methods often heuristically combine pretrained knowledge and few-shot evidence, we seek a more principled understanding of their relationship. In this paper, we propose a Bayesian-inspired optimal integration framework(BOIF) that interprets pretrained models as priors and few-shot evidence as likelihoods. Under conditional independence approximation, we show that the optimal log-posterior decomposes into the sum of prior logits and likelihood logits. This leads to a simple yet effective design principle: decouple the prior and likelihood pathways and combine their logits additively. Guided by this principle, we implement BOIF using CLIP with two novel enhancements: (1) a multi-level feature adapter to enrich visual representations, and (2) a simplified cache module for likelihood estimation. Extensive experiments on 11 benchmarks show BOIF achieves state-of-the-art performance (e.g., 80.61\% average accuracy at 16-shot) and strong out-of-distribution robustness. Our work provides both a principled perspective and an effective instantiation for few-shot adaptation.

We investigat the transferablility of pretained classification models. We wondered whether the ability of pre-trained models could be transferred to different types of tasks using only a handful of new examples. We took the pre‑trained components from the vision‑language model CLIP and combined them with a small set of labeled examples in a new way. By viewing the pre‑trained model as a source of general knowledge and the few new images as specific evidence, we derived a simple rule: just add the two sources of information together. We show that the method outperforms existing methods on 11 different datasets and handles unfamiliar image styles with ease. We have released a free and easy‑to‑use framework, called BOIF, along with the code. This lets anyone take a pre‑trained vision model, add a few labeled examples, and quickly adapt it to their own image‑classification problems.