Papers › Mixture of Experts Meets Decoupled Message Passing: Towards General and Adaptive Node...
Mixture of Experts Meets Decoupled Message Passing: Towards General and Adaptive Node Classification
Xuanze Chen, Jiajun Zhou, Shanqing Yu, Qi Xuan
Graph neural networks excel at graph representation learning but struggle with heterophilous data and long-range dependencies. And graph transformers address these issues through self-attention, yet face scalability and noise challenges on large-scale graphs. To overcome these limitations, we propose GNNMoE, a universal model architecture for node classification. This architecture flexibly combines fine-grained message-passing operations with a mixture-of-experts mechanism to build feature encoding blocks. Furthermore, by incorporating soft and hard gating layers to assign the most suitable expert networks to each node, we enhance the model's expressive power and adaptability to different graph types. In addition, we introduce adaptive residual connections and an enhanced FFN module into GNNMoE, further improving the expressiveness of node representation. Extensive experimental results demonstrate that GNNMoE performs exceptionally well across various types of graph data, effectively alleviating the over-smoothing issue and global noise, enhancing model robustness and adaptability, while also ensuring computational efficiency on large-scale graphs.
Code
Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.
Code Syntology ran Syntology
Not run by Syntology. Nothing on this page verifies that the listed code works.
Tasks
Results from the paper archive 2025-07-28
| Task | Dataset | Model | Metric | Value | Rank at snapshot | Leaderboard | Report |
|---|---|---|---|---|---|---|---|
| Node Classification | Actor | GNNMoE(SAGE-like P) | Accuracy | 37.97±1.01 | #11 of 62 | Archive leaderboard | report |
| Node Classification | Actor | GNNMoE(GAT-like P) | Accuracy | 37.76±0.98 | #17 of 62 | Archive leaderboard | report |
| Node Classification | Actor | GNNMoE(GCN-like P) | Accuracy | 37.59±1.36 | #20 of 62 | Archive leaderboard | report |
| Node Classification | Amazon Computers | GNNMoE(GCN-like P) | Accuracy | 92.17±0.50 | #4 of 12 | Archive leaderboard | report |
| Node Classification | Amazon Computers | GNNMoE(GAT-like P) | Accuracy | 91.98±0.46 | #5 of 12 | Archive leaderboard | report |
| Node Classification | Amazon Computers | GNNMoE(SAGE-like P) | Accuracy | 91.85±0.39 | #6 of 12 | Archive leaderboard | report |
| Node Classification | Amazon Photo | GNNMoE(GCN-like P) | Accuracy | 95.81±0.41 | #4 of 11 | Archive leaderboard | report |
| Node Classification | Amazon Photo | GNNMoE(GAT-like P) | Accuracy | 95.71±0.37 | #5 of 11 | Archive leaderboard | report |
| Node Classification | Amazon Photo | GNNMoE(SAGE-like P) | Accuracy | 95.46±0.24 | #6 of 11 | Archive leaderboard | report |
| Node Classification | Chameleon (48%/32%/20% fixed splits) | GNNMoE(GCN-like P) | Accuracy | 47.19±2.93 | #2 of 4 | Archive leaderboard | report |
| Node Classification | Chameleon (48%/32%/20% fixed splits) | GNNMoE(SAGE-like P) | Accuracy | 45.73±3.19 | #3 of 4 | Archive leaderboard | report |
| Node Classification | Chameleon (48%/32%/20% fixed splits) | GNNMoE(GAT-like P) | Accuracy | 45.56±3.94 | #4 of 4 | Archive leaderboard | report |
| Node Classification | Coauthor CS | GNNMoE(GCN-like P) | Accuracy | 95.81±0.26 | #5 of 24 | Archive leaderboard | report |
| Node Classification | Coauthor CS | GNNMoE(GAT-like P) | Accuracy | 95.72±0.23 | #8 of 24 | Archive leaderboard | report |
| Node Classification | Coauthor CS | GNNMoE(SAGE-like P) | Accuracy | 95.68±0.24 | #10 of 24 | Archive leaderboard | report |
| Node Classification | Coauthor Physics | GNNMoE(GAT-like P) | Accuracy | 97.05±0.19 | #6 of 14 | Archive leaderboard | report |
| Node Classification | Coauthor Physics | GNNMoE(GCN-like P) | Accuracy | 97.03±0.13 | #7 of 14 | Archive leaderboard | report |
| Node Classification | Coauthor Physics | GNNMoE(SAGE-like P) | Accuracy | 96.81±0.22 | #12 of 14 | Archive leaderboard | report |
| Node Classification | GNNMoE(GCN-like P) | Accuracy | 95.53±0.35 | #1 of 8 | Archive leaderboard | report | |
| Node Classification | GNNMoE(GAT-like P) | Accuracy | 95.21±0.25 | #2 of 8 | Archive leaderboard | report | |
| Node Classification | GNNMoE(SAGE-like P) | Accuracy | 94.63±0.36 | #3 of 8 | Archive leaderboard | report | |
| Node Classification | Penn94 | GNNMoE(GCN-like P) | Accuracy | 85.11±0.39 | #6 of 32 | Archive leaderboard | report |
| Node Classification | Penn94 | GNNMoE(SAGE-like P) | Accuracy | 84.05±0.37 | #12 of 32 | Archive leaderboard | report |
| Node Classification | Penn94 | GNNMoE(GAT-like P) | Accuracy | 81.98±0.47 | #16 of 32 | Archive leaderboard | report |
| Node Classification | Squirrel (48%/32%/20% fixed splits) | GNNMoE(GCN-like P) | Accuracy | 44.02±2.59 | #2 of 4 | Archive leaderboard | report |
| Node Classification | Squirrel (48%/32%/20% fixed splits) | GNNMoE(SAGE-like P) | Accuracy | 39.19±2.84 | #3 of 4 | Archive leaderboard | report |
| Node Classification | Squirrel (48%/32%/20% fixed splits) | GNNMoE(GAT-like P) | Accuracy | 39.19±3.94 | #4 of 4 | Archive leaderboard | report |
| Node Classification | roman-empire | GNNMoE(GAT-like P) | Accuracy (% ) | 87.29±0.60 | #5 of 7 | Archive leaderboard | report |
| Node Classification | roman-empire | GNNMoE(SAGE-like P) | Accuracy (% ) | 86.00±0.45 | #6 of 7 | Archive leaderboard | report |
| Node Classification | roman-empire | GNNMoE(GCN-like P) | Accuracy (% ) | 85.05±0.55 | #7 of 7 | Archive leaderboard | report |
| Node Classification | tolokers | GNNMoE(GAT-like P) | AUCROC | 85.45±0.94 | #2 of 4 | Archive leaderboard | report |
| Node Classification | tolokers | GNNMoE(GCN-like P) | AUCROC | 84.77±0.93 | #3 of 4 | Archive leaderboard | report |
| Node Classification | tolokers | GNNMoE(SAGE-like P) | AUCROC | 83.96±0.75 | #4 of 4 | Archive leaderboard | report |
| Node Property Prediction | ogbn-arxiv | GNNMoE(GAT-like P) | Test Accuracy | 0.7245±0.0032 | #60 of 86 | Archive leaderboard | report |
| Node Property Prediction | ogbn-arxiv | GNNMoE(GCN-like P) | Test Accuracy | 0.7229±0.0016 | #61 of 86 | Archive leaderboard | report |
| Node Property Prediction | ogbn-arxiv | GNNMoE(SAGE-like P) | Test Accuracy | 0.7194±0.0025 | #73 of 86 | Archive leaderboard | report |
Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections