Research map / Graph learning / Knowledge and temporal graphs
Logical query answering and rule reasoning
Answer logical queries and induce rules with neuro-symbolic path-based reasoning.
68 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2016–2026) · an approach in Knowledge and temporal graphs, keyword cluster “knowledge graph · entities · kg”
State History Benchmarks Tried Baselines Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 2 of 6 tracked benchmarks.
24 papers in 2023–24 and 18 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 9.3% to 6.4%.
Open questions the data points to
- YAGO3-10 (MRR): best went from 0.615 (up to 2024) to 0.631 in 2025 — DuetGraph: Coarse-to-Fine Knowledge Graph Reasoning with Dual-Pathway Global-Local Fusion (NeurIPS 2025 · 1 citation)
- FB15k-237 (MRR): best went from 0.43 (up to 2024) to 0.453 in 2025 — DuetGraph: Coarse-to-Fine Knowledge Graph Reasoning with Dual-Pathway Global-Local Fusion (NeurIPS 2025 · 1 citation)
- Papers report ConvE on FB15k-237 (MRR) anywhere from 0.325 to 0.41: comparisons against it depend on whose number you take.
What looks settled
- WN18RR (MRR): best up to 2024 0.67, best in 2025–26 0.593 from 2 papers, at most 0.2% better, if at all.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2016 · Column-Oriented Datalog Materialization for Large Knowledge Graphs (AAAI 2016 · 51 citations)
- 2018 · Go for a Walk and Arrive at the Answer: Reasoning Over Paths in Knowledge Bases using Reinforcement Learning (ICLR 2018 · 398 citations)
- 2019 · Probabilistic Logic Neural Networks for Reasoning (NeurIPS 2019 · 97 citations)
- 2020 · Graph-Based Reasoning over Heterogeneous External Knowledge for Commonsense Question Answering (AAAI 2020 · 191 citations)
- 2021 · Communicative Message Passing for Inductive Relation Reasoning (AAAI 2021 · 101 citations)
- 2022 · Fuzzy Logic Based Logical Query Answering on Knowledge Graphs (AAAI 2022 · 67 citations)
- 2023 · Neural Compositional Rule Learning for Knowledge Graph Reasoning (ICLR 2023 · 15 citations)
- 2024 · MINES: Message Intercommunication for Inductive Relation Reasoning over Neighbor-Enhanced Subgraphs (AAAI 2024 · 34 citations)
- 2025 · Towards Graph Foundation Models: Training on Knowledge Graphs Enables Transferability to General Graphs (NeurIPS 2025 · 3 citations)
- 2026 · NumCoKE: Ordinal-Aware Numerical Reasoning over Knowledge Graphs with Mixture-of-Experts and Contrastive Learning (AAAI 2026 · 1 citation)
Earliest papers here
- Column-Oriented Datalog Materialization for Large Knowledge Graphs (AAAI 2016 · 51 citations)
- Go for a Walk and Arrive at the Answer: Reasoning Over Paths in Knowledge Bases using Reinforcement Learning (ICLR 2018 · 398 citations)
- Knowledge Graph Embedding With Iterative Guidance From Soft Rules (AAAI 2018 · 215 citations)
Benchmark records
The best number each year that beat every earlier one, among papers of this approach reporting their own method. Numbers are extracted by AI from each paper's text; check the paper before citing.
WN18RR · MRR (higher is better)
10 papers, 2018–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2018 | 0.448 | Go for a Walk and Arrive at the Answer: Reasoning Over Paths in Knowledge Bases using Reinforcement Learning (ICLR 2018 · 398 citations) |
| 2023 | 0.67 | Neural Compositional Rule Learning for Knowledge Graph Reasoning (ICLR 2023 · 15 citations) |
YAGO3-10 · MRR (higher is better)
7 papers, 2023–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2023 | 0.556 | A*Net: A Scalable Path-based Reasoning Approach for Knowledge Graphs (NeurIPS 2023 · 10 citations) |
| 2024 | 0.615 | KnowFormer: Revisiting Transformers for Knowledge Graph Reasoning (ICML 2024 · 2 citations) |
| 2025 | 0.631 | DuetGraph: Coarse-to-Fine Knowledge Graph Reasoning with Dual-Pathway Global-Local Fusion (NeurIPS 2025 · 1 citation) |
FB15k-237 · MRR (higher is better)
6 papers, 2018–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2018 | 0.293 | Go for a Walk and Arrive at the Answer: Reasoning Over Paths in Knowledge Bases using Reinforcement Learning (ICLR 2018 · 398 citations) |
| 2019 | 0.332 | Probabilistic Logic Neural Networks for Reasoning (NeurIPS 2019 · 97 citations) |
| 2024 | 0.43 | KnowFormer: Revisiting Transformers for Knowledge Graph Reasoning (ICML 2024 · 2 citations) |
| 2025 | 0.453 | DuetGraph: Coarse-to-Fine Knowledge Graph Reasoning with Dual-Pathway Global-Local Fusion (NeurIPS 2025 · 1 citation) |
UMLS · MRR (higher is better)
6 papers, 2018–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2018 | 0.825 | Go for a Walk and Arrive at the Answer: Reasoning Over Paths in Knowledge Bases using Reinforcement Learning (ICLR 2018 · 398 citations) |
| 2023 | 0.869 | Rule Induction in Knowledge Graphs Using Linear Programming (AAAI 2023 · 3 citations) |
| 2024 | 0.971 | KnowFormer: Revisiting Transformers for Knowledge Graph Reasoning (ICML 2024 · 2 citations) |
Kinship · MRR (higher is better)
5 papers, 2018–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2018 | 0.72 | Go for a Walk and Arrive at the Answer: Reasoning Over Paths in Knowledge Bases using Reinforcement Learning (ICLR 2018 · 398 citations) |
| 2023 | 0.746 | Rule Induction in Knowledge Graphs Using Linear Programming (AAAI 2023 · 3 citations) |
NELL-995 · MRR (higher is better)
4 papers, 2018–2025 · one report since 2024, not better
| Year | Best so far | Set by |
|---|---|---|
| 2018 | 0.725 | Go for a Walk and Arrive at the Answer: Reasoning Over Paths in Knowledge Bases using Reinforcement Learning (ICLR 2018 · 398 citations) |
What has been tried
52 papers here report 553 measured changes (a component added, removed or swapped, compared with a baseline): helped 68%, no clear effect 7%, hurt 22%, mixed 3%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Add message intercommunication (M.I.) to GraIL | WN18RR · AUC-PR | 94.32 -> 99.50 | Helped | MINES: Message Intercommunication for Inductive Relation Reasoning over Neighbor-Enhanced Subgraphs (AAAI 2024 · 34 citations) |
| Add nested fact embedding loss (λ1=0.5) | FBHE · MRR | 0.368 -> 0.365 | Hurt | NestE: Modeling Nested Relational Structures for Knowledge Graph Reasoning (AAAI 2024 · 12 citations) |
| add prompt graph generation and encoding | Average (43 KGs) · MRR | 0.132 -> 0.442 | Helped | A Prompt-Based Knowledge Graph Foundation Model for Universal In-Context Reasoning (NeurIPS 2024 · 12 citations) |
| Remove edge types from the relation graph | Total average, 54 graphs · MRR | 0.366 -> 0.316 | Hurt | Towards Foundation Models for Knowledge Graph Reasoning (ICLR 2024 · 8 citations) |
| fine-tune the projection operator on complex queries rather than freeze the pretrained operator with thresholding | 23 datasets, EPFO avg · MRR | 0.279 -> 0.309 | Helped | A Foundation Model for Zero-shot Logical Query Reasoning (NeurIPS 2024 · 4 citations) |
| Add semantic-injected entity initialization (INIT2) | Inductive KG reasoning, node classification, and graph classification · performance | Helped | Towards Graph Foundation Models: Training on Knowledge Graphs Enables Transferability to General Graphs (NeurIPS 2025 · 3 citations) | |
| add rule guidance | FB15K-237 v1–v4 · F1 score | 88.6 -> 90.1; 93.7 -> 94.4; 93.8 -> 94.7; 94.5 -> 94.5 | Mixed | Rule-Guided Graph Neural Networks for Explainable Knowledge Graph Reasoning (AAAI 2025 · 2 citations) |
| add the proposed attention mechanism | FB15k-237 · MRR | 0.417 -> 0.430 | Helped | KnowFormer: Revisiting Transformers for Knowledge Graph Reasoning (ICML 2024 · 2 citations) |
| add the local pathway | FB15k-237, WN18RR, NELL-995, and YAGO3-10 · MRR | 0.445 -> 0.453; 0.584 -> 0.593; 0.582 -> 0.590; 0.617 -> 0.631 | Helped | DuetGraph: Coarse-to-Fine Knowledge Graph Reasoning with Dual-Pathway Global-Local Fusion (NeurIPS 2025 · 1 citation) |
| use 8 layers instead of 6 layers, with 10% of entities | WN18RR · MRR | 0.566 -> 0.567 | Helped | Less is More: One-shot Subgraph Reasoning on Large-scale Knowledge Graphs (ICLR 2024 · 1 citation) |
| Add 3-path motifs instead of using 2-path motifs | 51 KGs, zero-shot, total average · MRR | 0.370 -> 0.378 | Helped | How Expressive are Knowledge Graph Foundation Models? (ICML 2025 · 0 citations) |
| Add PCRA filtering to CTP3 | WN18RR · MRR | 0.37 -> 0.39 | Helped | Improving Soft Unification with Knowledge Graph Embedding Methods (ICML 2025) |
Baseline numbers that disagree
The same baseline on the same benchmark and metric, reported more than 5% apart by different papers here. Training budgets, data splits and reimplementations differ, so state which number you compare with.
| Baseline | Benchmark | Metric | Reported | Papers |
|---|---|---|---|---|
| ConvE | FB15k-237 | MRR | 0.325 – 0.41 | 2 papers, e.g. NeurIPS 2019 and ICLR 2018 |
| DRUM | Family | MRR | 0.89 – 0.958 | 2 papers, e.g. ICLR 2023 and ICLR 2023 |
| RNNLogic | Kinship | MRR | 0.639 – 0.687 | 2 papers, e.g. ICLR 2023 and AAAI 2023 |
What reviewers will expect
Datasets and baselines named in the 11 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- WN18RR · 4 of 11 papers
- FB15k-237 · 4 of 11 papers
- FB15k · 2 of 11 papers
- UMLS · 2 of 11 papers
- CLUTRR · 2 of 11 papers
- CoDEx Medium · 2 of 11 papers
- NELL-995 · 2 of 11 papers
Baselines
- DRUM · 4 of 11 papers
- NBFNet · 4 of 11 papers
- NCRL · 2 of 11 papers
- GNTP · 2 of 11 papers
- CTP · 2 of 11 papers
- NeuralLP · 2 of 11 papers
- ULTRA · 2 of 11 papers
- InGram · 2 of 11 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 42 papers; from OpenAlex and the papers' affiliations).
- Intel · 4 papers
- University of Oxford · 3 papers
- Hong Kong University of Science and Technology · 3 papers
- Hong Kong Baptist University · 3 papers
- Mila · 3 papers
- Université de Montréal · 3 papers
- Beihang University · 2 papers
- Beijing University of Posts and Telecommunications · 2 papers
- Korea University · 2 papers
- Griffith University · 2 papers
Use it
Download BibTeX of 39 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Knowledge graph embeddings (110 papers)
- Temporal knowledge graph reasoning (84 papers)
- Inductive knowledge graph completion (38 papers)
Elsewhere in the topic
- Continuous-time dynamic graph learning (132 papers)
- Graph reasoning for visual recognition (96 papers)
- Graph-based dialog systems (94 papers)
- Social and behavior graphs (87 papers)
- Scene graph generation (70 papers)
- Temporal graph neural architectures (69 papers)
- Heterogeneous graph learning (66 papers)
- Spatio-temporal graph traffic forecasting (60 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).