Cross Lingual Question Answering On Tydiqa
評価指標
EM
評価結果
このベンチマークにおける各モデルのパフォーマンス結果
モデル名 | EM | Paper Title | Repository |
|---|---|---|---|
| U-PaLM-540B (CoT) | 54.6 | Transcending Scaling Laws with 0.1% Extra Compute | - |
| Flan-PaLM 540B (direct-prompting) | 67.8 | Scaling Instruction-Finetuned Language Models | |
| Decoupled | 42.8 | Rethinking embedding coupling in pre-trained language models | |
| ByT5 XXL | 60.0 | ByT5: Towards a token-free future with pre-trained byte-to-byte models | |
| ByT5 (fine-tuned) | 81.9 | ByT5: Towards a token-free future with pre-trained byte-to-byte models | |
| PaLM 2-M (one-shot) | - | PaLM 2 Technical Report | |
| Flan-U-PaLM 540B (direct-prompting) | 68.3 | Scaling Instruction-Finetuned Language Models | |
| PaLM 2-S (one-shot) | - | PaLM 2 Technical Report | |
| U-PaLM 62B (fine-tuned) | 78.4 | Transcending Scaling Laws with 0.1% Extra Compute | - |
| PaLM 2-L (one-shot) | - | PaLM 2 Technical Report | |
| PaLM-540B (CoT) | 52.9 | PaLM: Scaling Language Modeling with Pathways |
0 of 11 row(s) selected.