Node Classification on PubMed (accuracy, Avg.Rank)
91AccuracyGraphICL-S1
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GraphICL-S1Backbone=Qwen2.5-32B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 91 | 2.2 | |
| GraphSearch-RBackbone=Qwen2.5-32B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 90.4 | 2.7 | |
| GraphSearch-FBackbone=Qwen2.5-72B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 90 | 2 | |
| GraphSearch-RBackbone=Qwen2.5-72B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 89.9 | 2.3 | |
| GraphSearch-FBackbone=Qwen2.5-32B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 89.8 | 1.8 | |
| Search-o1Backbone=Qwen2.5-32B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 89.1 | 5 | |
| Search-o1Backbone=Qwen2.5-72B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 88.8 | 4.3 | |
| GraphICL-S1Backbone=Qwen2.5-72B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 87.4 | 2.3 | |
| Chain-of-ThoughtBackbone=Qwen2.5-32B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 86.6 | 5.7 | |
| GraphICL-S2Backbone=Qwen2.5-32B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 86.3 | 4.2 | |
| Chain-of-ThoughtBackbone=Qwen2.5-72B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 80.3 | 6 | |
| GraphICL-S2Backbone=Qwen2.5-72B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 79.3 | 4.3 | |
| Graph-CoTBackbone=Qwen2.5-32B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 70.5 | 6.5 | |
| Graph-CoTBackbone=Qwen2.5-72B-Instruct, Evaluation Protocol=Zero-shot2026.01 | 67.8 | 6.7 |