CHoE: Cross-Domain Heterogeneous Graph Prompt Learning via Structure-Conditioned Experts
The article discusses a new method called CHoE for Cross-Domain Heterogeneous Graph Prompt Learning. This approach aims to enhance the performance of pre-trained models in diverse application domains. CHoE utilizes structure-conditioned experts and a prompt-based semantic fusion module to improve predictions in few-shot cross-domain scenarios.
- ▪Heterogeneous Graph Prompt Learning (HGPL) is designed to bridge the gap between pre-training models and their applications.
- ▪Current HGPL methods are limited to in-domain scenarios, which affects their performance in real-world applications.
- ▪CHoE introduces structure-conditioned experts and a routing mechanism to select compatible experts for each meta-path view.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Machine Learning arXiv:2605.15888 (cs) [Submitted on 15 May 2026] Title:CHoE: Cross-Domain Heterogeneous Graph Prompt Learning via Structure-Conditioned Experts Authors:Peiyuan Li, Yongqi Huang, Jitao Zhao, Dongxiao He, Di Jin, Weixiong Zhang View a PDF of the paper titled CHoE: Cross-Domain Heterogeneous Graph Prompt Learning via Structure-Conditioned Experts, by Peiyuan Li and 5 other authors View PDF HTML (experimental) Abstract:Heterogeneous Graph Prompt Learning (HGPL)has emerged as a promising paradigm for bridging the gap between the objectives of pre-training foundation models and their downstream applications in heterogeneous graph settings.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
