Research map / Graph learning / Knowledge and temporal graphs
Temporal knowledge graph reasoning
Forecast and reason over time-evolving knowledge graphs using rules and sequence models.
84 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 Checklist Preprints Who Export
Where it stands
Still moving The best reported result is still improving on 1 of 3 tracked benchmarks.
26 papers in 2023–24 and 22 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 10.1% to 7.8%.
Open questions the data points to
- ICEWS14 (MRR): best went from 62.61 (up to 2024) to 90.7 in 2025 — TeDS: Joint Learning of Diachronic and Synchronic Perspectives in Quaternion Space for Temporal Knowledge Graph Completion (ICML 2025)
What looks settled
- ICEWS18 (MRR): best up to 2024 63.3, best in 2025–26 47.02 from 2 papers, at most 0.2% better, if at all.
- WIKI (MRR): best up to 2024 99.67, best in 2025–26 83.21 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 · A Joint Model for Entity Set Expansion and Attribute Extraction from Web Search Queries (AAAI 2016 · 12 citations)
- 2017 · Know-Evolve: Deep Temporal Reasoning for Dynamic Knowledge Graphs (ICML 2017 · 106 citations)
- 2018 · Does William Shakespeare REALLY Write Hamlet? Knowledge Representation Learning With Confidence (AAAI 2018 · 84 citations)
- 2019 · Explainable Reasoning over Knowledge Graphs for Recommendation (AAAI 2019 · 743 citations)
- 2020 · CASIE: Extracting Cybersecurity Event Information from Text (AAAI 2020 · 113 citations)
- 2021 · Learning from History: Modeling Temporal Knowledge Graphs with Sequential Copy-Generation Networks (AAAI 2021 · 320 citations)
- 2022 · TLogic: Temporal Logical Rules for Explainable Link Forecasting on Temporal Knowledge Graphs (AAAI 2022 · 169 citations)
- 2023 · Temporal Knowledge Graph Reasoning with Historical Contrastive Learning (AAAI 2023 · 159 citations)
- 2024 · MKG-FENN: A Multimodal Knowledge Graph Fused End-to-End Neural Network for Accurate Drug–Drug Interaction Prediction (AAAI 2024 · 41 citations)
- 2025 · CognTKE: A Cognitive Temporal Knowledge Extrapolation Framework (AAAI 2025 · 9 citations)
- 2026 · GraphRAG-Induced Dual Knowledge Structure Graphs for Personalized Learning Path Recommendation (AAAI 2026 · 5 citations)
Earliest papers here
- A Joint Model for Entity Set Expansion and Attribute Extraction from Web Search Queries (AAAI 2016 · 12 citations)
- Know-Evolve: Deep Temporal Reasoning for Dynamic Knowledge Graphs (ICML 2017 · 106 citations)
- Radon – Rapid Discovery of Topological Relations (AAAI 2017 · 55 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.
ICEWS14 · MRR (higher is better)
5 papers, 2021–2025 · still improving
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 62.53 | ChronoR: Rotation Based Temporal Knowledge Graph Embedding (AAAI 2021 · 118 citations) |
| 2024 | 62.61 | DECRL: A Deep Evolutionary Clustering Jointed Temporal Knowledge Graph Representation Learning Approach (NeurIPS 2024 · 6 citations) |
| 2025 | 90.7 | TeDS: Joint Learning of Diachronic and Synchronic Perspectives in Quaternion Space for Temporal Knowledge Graph Completion (ICML 2025) |
ICEWS18 · MRR (higher is better)
5 papers, 2021–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 46.69 | Learning from History: Modeling Temporal Knowledge Graphs with Sequential Copy-Generation Networks (AAAI 2021 · 320 citations) |
| 2024 | 63.3 | DECRL: A Deep Evolutionary Clustering Jointed Temporal Knowledge Graph Representation Learning Approach (NeurIPS 2024 · 6 citations) |
WIKI · MRR (higher is better)
4 papers, 2021–2025 · flat since 2024
| Year | Best so far | Set by |
|---|---|---|
| 2021 | 45.5 | Learning from History: Modeling Temporal Knowledge Graphs with Sequential Copy-Generation Networks (AAAI 2021 · 320 citations) |
| 2024 | 99.67 | DECRL: A Deep Evolutionary Clustering Jointed Temporal Knowledge Graph Representation Learning Approach (NeurIPS 2024 · 6 citations) |
What has been tried
64 papers here report 645 measured changes (a component added, removed or swapped, compared with a baseline): helped 75%, no clear effect 4%, hurt 17%, mixed 4%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use P1 channel only instead of all four channels | DDI Task 1 (known drugs) · ACC | 0.9409 -> 0.9172 | Hurt | MKG-FENN: A Multimodal Knowledge Graph Fused End-to-End Neural Network for Accurate Drug–Drug Interaction Prediction (AAAI 2024 · 41 citations) |
| Use duration prediction (-td) with event-split modelling | YAGO11k · aeIOU | 0.2996 -> 0.2675 | Hurt | TEILP: Time Prediction over Knowledge Graphs via Logical Reasoning (AAAI 2024 · 23 citations) |
| add adaptive confidence | FB15K-237, top-1% · precision@top-K | .951 -> .969 | Helped | Knowledge Graph Error Detection with Contrastive Confidence Adaption (AAAI 2024 · 18 citations) |
| Add temporal order encoding to timestamp embeddings | CronQuestions overall · Hits@1 | 0.902 -> 0.971 | Helped | Question Calibration and Multi-Hop Modeling for Temporal Question Answering (AAAI 2024 · 16 citations) |
| add random noise image instead of matched image | DL · MRR / w-F1 | MRR 22.64 to 18.62; w-F1 44.00 to 27.23 | Hurt | Beyond Entities: A Large-Scale Multi-Modal Knowledge Graph with Triplet Fact Grounding (AAAI 2024 · 10 citations) |
| add node-level attention (α) | MIMIC-III drug recommendation · F1-score | 64.6 -> 66.8 | Helped | GraphCare: Enhancing Healthcare Predictions with Personalized Knowledge Graphs (ICLR 2024 · 10 citations) |
| add the global shallow reasoner (System1) | ICE14, ICE18, and ICE05-15 · MRR | 36.12 -> 46.06; 29.01 -> 35.24; 40.61 -> 53.13 | Helped | CognTKE: A Cognitive Temporal Knowledge Extrapolation Framework (AAAI 2025 · 9 citations) |
| Use GNN-based GFM retrieval instead of pre-trained or fine-tuned all-mpnet-v2 embeddings | HotpotQA · R@2 | 59.4 -> 78.3; 67.0 -> 78.3 | Helped | GFM-RAG: Graph Foundation Model for Retrieval Augmented Generation (NeurIPS 2025 · 8 citations) |
| Add unsupervised cluster alignment | ICEWS14 · MRR | 59.16 (-3.5) -> 62.61 | Helped | DECRL: A Deep Evolutionary Clustering Jointed Temporal Knowledge Graph Representation Learning Approach (NeurIPS 2024 · 6 citations) |
| Use core concepts (SCC) instead of Top_10 as the concept reference | ADE20k false prediction explanation · CM^FP (%) | 42.52 -> 78.03 | Helped | Knowledge-Aware Neuron Interpretation for Scene Classification (AAAI 2024 · 5 citations) |
| Remove ER-N/F (equivalent to adding ER-N/F) | NER evaluation · F1 | 92.96 -> 92.96 | No clear effect | Construction and Application of Materials Knowledge Graph in Multidisciplinary Materials Science via Large Language Model (NeurIPS 2024 · 5 citations) |
| Add data augmentation | HyperRED hyper-relational / Test · F1 | 72.06 ± 0.34 -> 83.63 ± 0.63 | Helped | Text2NKG: Fine-Grained N-ary Relation Extraction for N-ary relational Knowledge Graph Construction (NeurIPS 2024 · 4 citations) |
What reviewers will expect
Datasets and baselines named in the 12 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Datasets
- WIKI · 2 of 12 papers
- ICEWS14 · 2 of 12 papers
- ICEWS18 · 2 of 12 papers
- HotpotQA · 2 of 12 papers
Baselines
- CyGNet · 2 of 12 papers
- CENET · 2 of 12 papers
- BM25 · 2 of 12 papers
- LightRAG · 2 of 12 papers
- GraphRAG · 2 of 12 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 48 papers; from OpenAlex and the papers' affiliations).
- Zhejiang University · 4 papers
- Beihang University · 3 papers
- Georgia Institute of Technology · 3 papers
- University of Electronic Science and Technology of China · 2 papers
- Peking University · 2 papers
- Nanjing University · 2 papers
- Nanyang Technological University · 2 papers
- Fudan University · 2 papers
- Beijing Jiaotong University · 2 papers
- University of Illinois at Urbana-Champaign · 2 papers
Use it
Download BibTeX of 40 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)
- Logical query answering and rule reasoning (68 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).