Knowledge Graph Extraction on HotpotQA
1.18Avg Edge MultiplicityIN-CONTEXT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| IN-CONTEXTBase Model=Qwen3-30B-A3B2026.05 | 1.18 | 14.1 | 515.7 | 1,821 | |
| WIKONTICBase Model=Qwen3-30B-A3B2026.05 | 1.05 | 4.3 | 339.7 | 895.3 | |
| WIKONTICBase Model=GPT-OSS-120B2026.05 | 1.05 | 4.4 | 370.3 | 1,068 | |
| OAKBase Model=Qwen3-30B-A3B2026.05 | 1.04 | 3.2 | 336 | 884.7 | |
| OAKBase Model=GPT-OSS-120B2026.05 | 1.04 | 3.2 | 329 | 823.3 | |
| OAK+MENDBase Model=Qwen3-30B-A3B2026.05 | 1.03 | 2.7 | 402.7 | 1,141.7 | |
| OAK+MENDBase Model=GPT-OSS-120B2026.05 | 1.03 | 2.6 | 383.3 | 1,054.7 | |
| IN-CONTEXTBase Model=GPT-OSS-120B2026.05 | 1.02 | 2 | 320.7 | 775.3 |