Temporal Knowledge Graph Question Answering on Timeline-ICEWS (test)
64.2Overall AccuracyTemp-R1
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Temp-R1Backbone=Llama3.1-8B-Instruct2026.01 | 64.2 | 86.6 | 38.8 | 59.5 | |
| PoK2026.01 | 60.2 | 74.4 | 45.6 | 57.8 | |
| TKG-Thinker (Qwen2.5-7B-Instruct)base_model=Qwen2.5-7B-Instruct, framework=TKG-Thinker2026.02 | 50.8 | 55.6 | 40.9 | 58.3 | |
| TKG-Thinker (Llama3-8B-Instruct)base_model=Llama3-8B-Instruct, framework=TKG-Thinker2026.02 | 46.2 | 59.6 | 27.3 | 53.3 | |
| RAG baseline2026.02 | 26.5 | 66 | 12.8 | 1.1 | |
| RAG Baseline2026.01 | 26.5 | 66 | 12.8 | 1.1 | |
| GPT-4o2026.02 | 11.3 | 5.1 | 3.5 | 35.3 | |
| GPT-4o2026.01 | 11.3 | 5.1 | 3.5 | 35.3 | |
| LLaMA2-7B2026.02 | 11.1 | 3.5 | 6.6 | 32.2 | |
| LLaMA2-7B2026.01 | 11.1 | 3.5 | 6.6 | 32.2 |