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

2026-04-17

AI Summary

This work addresses the “domain gravity” problem in cross-disciplinary few-shot class-incremental learning, where domain heterogeneity and data imbalance induce prototype drift, disproportionately degrading performance on underrepresented or high-entropy domains. The study formally defines this phenomenon and introduces XD-VSCIL, a more realistic benchmark that better reflects practical cross-domain scenarios. To mitigate these issues, the authors propose HyCal, a training-free method that operates in the frozen CLIP feature space by fusing cosine similarity with Mahalanobis distance. This hybrid approach simultaneously captures directional alignment and covariance-aware magnitude information to calibrate class prototypes effectively. HyCal significantly outperforms existing methods, achieving strong performance while preserving knowledge of previously seen classes, and offers an advantageous trade-off between computational efficiency and accuracy.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Adversarial Attacks & Robustness
• Application Domains: Humanities & Computational Social Science

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

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.

Problem

• Few-Shot Class-Incremental Learning
• Cross-Discipline
• Domain Gravity
• Data Imbalance
• Heterogeneous Domains

Innovation

• Domain Gravity
• Prototype Calibration
• Few-Shot Class-Incremental Learning
• Vision-Language Models
• Training-Free Method