Continual learning for smart contract vulnerability detection

Investigate the application of continual learning to smart contract vulnerability detection, particularly for sequential multi-label vulnerability tasks with evolving vulnerability categories and strong semantic correlations.

Background

The paper observes that continual learning has been studied extensively in computer vision and natural language processing, but its use for smart contract vulnerability detection has received little investigation. The problem is complicated by the sequential emergence of vulnerability categories, the multi-label nature of vulnerability detection, and semantic correlations among vulnerability types, all of which can intensify catastrophic forgetting.

References

Whether this advantage extends to multi-label audio CIL, where labels cover only the current class group and leave old sound events unannotated, remains an open question.

— Exemplar-Free Analytic Learning for Multi-Label Audio Class-Incremental Learning  (2609.29777 - Luo et al., 24 Sep 2026) in Section 1, Introduction

Although continual learning has achieved considerable success in computer vision and natural language processing, its application to smart contract vulnerability detection remains largely unexplored.

— Frequency-Aware Continual Learning for Smart Contract Vulnerability Detection with Large Language Models  (2608.19680 - Huang et al., 20 Aug 2026) in Section 2, subsection “Continual Learning for Sequential Tasks”