Advancing Analytic Class-Incremental Learning through Vision-Language Calibration

2026-02-14

AI Summary

This work addresses the challenge in class-incremental learning where pretrained models suffer from representational rigidity, causing analytic learning to be vulnerable to cumulative errors and feature incompatibility, thereby struggling to balance adaptation efficiency with long-term stability. To overcome this, the paper proposes VILA, a dual-branch framework that systematically uncovers, for the first time, the failure mechanisms of analytic class-incremental learning. It introduces a two-level calibration strategy: at the feature level, it fuses task-adaptive features with frozen semantic anchors; at the decision level, it leverages cross-modal semantic priors to correct prediction bias. This approach effectively mitigates representational rigidity and achieves state-of-the-art performance across eight benchmarks, notably unifying high accuracy and efficiency in fine-grained and long-sequence scenarios.

Technology Category

Computer Vision: Representation Learning for VisionMachine Learning: Calibration & Uncertainty QuantificationSearch and Optimization: Learning to Search

Application Category

Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingSemantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMsGraph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Class-incremental learning (CIL) with pre-trained models (PTMs) faces a critical trade-off between efficient adaptation and long-term stability. While analytic learning enables rapid, recursive closed-form updates, its efficacy is often compromised by accumulated errors and feature incompatibility. In this paper, we first conduct a systematic study to dissect the failure modes of PTM-based analytic CIL, identifying representation rigidity as the primary bottleneck. Motivated by these insights, we propose \textbf{VILA}, a novel dual-branch framework that advances analytic CIL via a two-level vision-language calibration strategy. Specifically, we coherently fuse plastic, task-adapted features with a frozen, universal semantic anchor at the feature level through geometric calibration, and leverage cross-modal priors at the decision level to rectify prediction bias. This confluence maintains analytic-learning's extreme efficiency while overcoming its inherent brittleness. Extensive experiments across eight benchmarks demonstrate that VILA consistently yields superior performance, particularly in fine-grained and long-sequence scenarios. Our framework harmonizes high-fidelity prediction with the simplicity of analytic learning. Our code is available at https://github.com/byzhaoAI/VILA

Problem

• Class-Incremental Learning
• Pre-trained Models
• Analytic Learning
• Representation Rigidity
• Feature Incompatibility

Innovation

• analytic class-incremental learning
• vision-language calibration
• representation rigidity
• dual-branch framework
• cross-modal priors