Question Answering on COMPLEXTR
82.2F1 ScoreSchema-Guided
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Schema-GuidedBackbone=GPT-4.12026.06 | 82.2 | 74.5 | 82 | 79.5 | |
| ProbTreeBackbone=GPT-4.12026.06 | 80.5 | 70 | 79.7 | 77 | |
| w/o struc. intel.Backbone=GPT-4.12026.06 | 78.6 | 67.7 | 77.9 | 77.2 | |
| Best publishedEvaluation Setting=zero-shot / no-finetuning2026.06 | 78.2 | — | 72.7 | 81.5 | |
| w/o KG ingestionBackbone=GPT-4.12026.06 | 77.1 | 68.8 | 76.8 | 74.5 | |
| w/o schema guidanceBackbone=GPT-4.12026.06 | 77 | 67.2 | 76.8 | 74.4 | |
| single-passBackbone=GPT-4.12026.06 | 76.3 | 65 | 75.7 | 74.4 | |
| AtomRBackbone=GPT-4.12026.06 | 75.7 | 64 | 74.7 | 74.5 | |
| RAGBackbone=GPT-4.12026.06 | 53.3 | 25.5 | 44.6 | 69 | |
| CoKBackbone=GPT-4.12026.06 | 40 | 28 | 38.6 | 44.5 |