Search papers, labs, and topics across Lattice.
This paper addresses the challenges of smart contract vulnerability detection using Large Language Models (LLMs) by proposing a three-stage pipeline that integrates Frequency-Aware Low-Rank Adaptation (FA-LoRA), Forget-Aware Replay (FAR), and Anchor-Protected Progressive Merging (APPM). The approach allows for parameter-efficient adaptation to new vulnerability categories while mitigating catastrophic forgetting and consolidating multiple adapters into a single model. The proposed framework achieves a Micro-F1 score of 0.8085, demonstrating its effectiveness in evolving blockchain ecosystems while maintaining low computational costs.
Achieving a Micro-F1 score of 0.8085, this framework effectively balances adaptation and retention in LLMs for smart contract vulnerability detection.
Smart contract vulnerability detection with Large Language Models (LLMs) faces three causally linked challenges. First, new vulnerability categories demand parameter-efficient adaptation, since full retraining is prohibitive for sequentially arriving tasks. Second, training per-task adapters on a shared backbone causes catastrophic forgetting of previously learned vulnerabilities. Third, the resulting multiplicity of adapters must be consolidated into a single model, since task identity is unknown at inference time. Each challenge arises directly from the solution to its predecessor, making an integrated framework essential. We propose a three-stage pipeline in which each stage addresses one challenge and feeds into the next. The adaptation stage uses Frequency-Aware Low-Rank Adaptation (FA-LoRA), which performs adaptation in the Fourier domain with per-frequency importance gates, requiring only 0.4% trainable parameters while outperforming standard LoRA and QLoRA. The continual learning stage applies Forget-Aware Replay (FAR), which uses these frequency gates to estimate per-sample forgetting risk via loss dynamics and prioritizes vulnerable knowledge for rehearsal, achieving an average Micro-F1 of 0.8022 across sequential tasks. The deployment stage employs Anchor-Protected Progressive Merging (APPM), which exploits the asymmetric generalization produced by FAR training to identify the strongest-generalizing adapter as an anchor and consolidates all adapters into a single model via anchor-protected weighted merging with frequency-domain gate competition. APPM achieves a Micro-F1 of 0.8085, within 2.7% of the independent per-task upper bound, at a merge cost of 156 ms and no additional runtime memory. Experiments on DIVE confirm the framework effectively addresses all three challenges for evolving blockchain ecosystems.
