Research map / Graph learning / GNN architectures
Efficient Probabilistic Logic Reasoning with Graph Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use ExpressGNN-E instead of the strongest reported baseline, HL-MRF | MLN inference · UW-CSE Graphics | AUC-PR: 0.06 -> 0.19 | Helped |
| Use ExpressGNN-E instead of the strongest reported baseline | MLN inference · Cora | AUC-PR | Helped |
| Use ExpressGNN-E instead of HL-MRF | MLN inference · Kinship S1–S4 | AUC-PR: 1.00 -> 0.97; 1.00 -> 0.97; 1.00 -> 0.99; 1.00 -> 0.99 | Hurt |
| Add tunable embeddings to GNN64 (Tune4) (ablation) | ExpressGNN-E · Cora S2 | AUC-PR: 0.58 -> 0.75 | Helped |
| Add tunable embeddings to GNN64 (Tune4) (ablation) | ExpressGNN-E · Cora S5 | AUC-PR: 0.53 -> 0.70 | Helped |
| Add tunable embeddings to GNN64 (Tune4) (ablation) | ExpressGNN-E · Cora S4 | AUC-PR: 0.54 -> 0.54 | No clear effect |
| Remove tunable embeddings and use GNN64 only (ablation) | ExpressGNN-E · Cora S2 | AUC-PR: 0.75 -> 0.58 | Hurt |
| Remove tunable embeddings and use GNN64 only (ablation) | ExpressGNN-E · Cora S5 | AUC-PR: 0.70 -> 0.53 | Hurt |
| Replace inference-only ExpressGNN-E with rule-weight learning (ExpressGNN-EM) (ablation) | ExpressGNN on FB15K-237 · FB15K-237 100% training | MRR: 0.45 -> 0.49 | Helped |
| Replace inference-only ExpressGNN-E with rule-weight learning (ExpressGNN-EM) (ablation) | ExpressGNN on FB15K-237 · FB15K-237 100% training | Hits@10: 57.3 -> 60.8 | Helped |
| Use ExpressGNN-EM rather than Neural LP | zero-shot relational learning · FB15K-237 zero-shot | MRR: 0.010 -> 0.185 | Helped |
| Use ExpressGNN-EM rather than Neural LP | zero-shot relational learning · FB15K-237 zero-shot | Hits@10: 2.7 -> 29.6 | Helped |
Closest papers by meaning
- Probabilistic Logic Neural Networks for Reasoning (NeurIPS 2019 · similarity 0.93)
- Rethinking and Extending the Probabilistic Inference Capacity of GNNs (ICLR 2024 · similarity 0.89)
- The Correspondence Between Bounded Graph Neural Networks and Fragments of First-Order Logic (AAAI 2026 · similarity 0.89)
- LogicMP: A Neuro-symbolic Approach for Encoding First-order Logic Constraints (ICLR 2024 · similarity 0.88)
- Systematic Relational Reasoning With Epistemic Graph Neural Networks (ICLR 2025 · similarity 0.87)
- Recurrent Graph Neural Networks and Their Connections to Bisimulation and Logic (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.