Precise detection of legal loopholes is a critical component in upholding judicial fairness.However, existing methods exhibit significant shortcomings in deeply integrating legal knowledge, while their modelling capabilities for imbalanced data remain underdeveloped.To address this, this paper first designs a hierarchical experience replay mechanism.By storing and sampling experiences through temporal difference errors and task priorities, it effectively controls gradient conflicts during cross-task training.Second, a legal loophole detection model based on improved multi-agent reinforcement learning is designed.The knowledge fusion module maps multi-source legal knowledge into a unified representation space, achieving selective knowledge enhancement through a learnable knowledge gating mechanism.Furthermore, an adaptive feature space partitioning is realised through the collaborative mechanism of multiple agents across multi-classification tasks, significantly improving the recognition performance of minority class samples.Experimental results demonstrate that the proposed model achieves a loophole detection accuracy of 92.48%, significantly enhancing detection precision.
Dongli Ma (Thu,) studied this question.