Multi-hop Knowledge-based Question Answering on PathQuestion-Large (PQL) 3H 1.0
95.4AccuracyHypergraph Transformer
Evaluation results
| Method | Links | |
|---|---|---|
| Hypergraph TransformerSupervision=Weakly-supervised, Note=PQL-3H-More data (2x QA pairs)2022.04 | 95.4 | |
| Hypergraph TransformerSupervision=Weakly-supervised2022.04 | 77.9 | |
| SRNSupervision=Weakly-supervised2022.04 | 77.5 | |
| IRNSupervision=Fully-supervised2022.04 | 71 | |
| MINERVASupervision=Weakly-supervised2022.04 | 65.7 | |
| Seq2SeqSupervision=Fully-supervised2022.04 | 64.7 | |
| KV-MemNNSupervision=Fully-supervised2022.04 | 63.4 | |
| IRN-weakSupervision=Weakly-supervised2022.04 | 61.8 | |
| MemNNSupervision=Fully-supervised2022.04 | 53.6 | |
| EmbedSupervision=Weakly-supervised2022.04 | 22.5 | |
| SubgraphSupervision=Weakly-supervised2022.04 | 21.3 |