Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

HyCal: A Training-Free Prototype Calibration Method for Cross-Discipline Few-Shot Class-Incremental Learning

About

Pretrained Vision-Language Models (VLMs) like CLIP show promise in continual learning, but existing Few-Shot Class-Incremental Learning (FSCIL) methods assume homogeneous domains and balanced data distributions, limiting real-world applicability where data arises from heterogeneous disciplines with imbalanced sample availability and varying visual complexity. We identify Domain Gravity, a representational asymmetry where data imbalance across heterogeneous domains causes overrepresented or low-entropy domains to disproportionately influence the embedding space, leading to prototype drift and degraded performance on underrepresented or high-entropy domains. To address this, we introduce Cross-Discipline Variable Few-Shot Class-Incremental Learning (XD-VSCIL), a benchmark capturing real-world heterogeneity and imbalance where Domain Gravity naturally intensifies. We propose Hybrid Prototype Calibration (HyCal), a training-free method combining cosine similarity and Mahalanobis distance to capture complementary geometric properties-directional alignment and covariance-aware magnitude-yielding stable prototypes under imbalanced heterogeneous conditions. Operating on frozen CLIP embeddings, HyCal achieves consistent retention-adaptation improvements while maintaining efficiency. Experiments show HyCal effectively mitigates Domain Gravity and outperforms existing methods in imbalanced cross-domain incremental learning.

Eunju Lee, MiHyeon Kim, JuneHyoung Kwon, Yoonji Lee, JiHyun Kim, Soojin Jang, YoungBin Kim• 2026

Related benchmarks

TaskDatasetResultRank
Class-incremental learningXD-VSCIL High-Scale Domain Imbalance
Aircraft Score40.98
11
Few-Shot Class-Incremental LearningXD-VSCIL 5-shot
Last Accuracy60.82
5
Few-Shot Class-Incremental LearningXD-VSCIL 10-shot
Last Accuracy66.52
5
Few-Shot Class-Incremental LearningXD-VSCIL 15-shot
Last Accuracy69.02
5
Few-Shot Class-Incremental LearningXD-VSCIL 20-shot
Last Accuracy70.67
5
Cross-Domain Variable-Shot Continual Incremental LearningXD-VSCIL Balanced-in-Class Domain
S_adapt53.63
5
Cross-Domain Variable-Shot Continual Incremental LearningXD-VSCIL Cross-Scale Imbalance
S_adapt51.99
5
Cross-Domain Variable-Shot Continual Incremental LearningXD-VSCIL High-Scale Domain Imbalance
S_adapt47.71
5
Showing 8 of 8 rows

Other info

Follow for update