| Paper | Code | EM | F1 | ModelName | ReleaseDate |
|---|---|---|---|---|---|
| ByT5: Towards a token-free future with pre-trained byte-to-byte models | ✓ Link | 81.9 | ByT5 (fine-tuned) | 2021-05-28 | |
| Transcending Scaling Laws with 0.1% Extra Compute | 78.4 | 88.5 | U-PaLM 62B (fine-tuned) | 2022-10-20 | |
| Scaling Instruction-Finetuned Language Models | ✓ Link | 68.3 | Flan-U-PaLM 540B (direct-prompting) | 2022-10-20 | |
| Scaling Instruction-Finetuned Language Models | ✓ Link | 67.8 | Flan-PaLM 540B (direct-prompting) | 2022-10-20 | |
| ByT5: Towards a token-free future with pre-trained byte-to-byte models | ✓ Link | 60.0 | 75.3 | ByT5 XXL | 2021-05-28 |
| Transcending Scaling Laws with 0.1% Extra Compute | 54.6 | U-PaLM-540B (CoT) | 2022-10-20 | ||
| PaLM: Scaling Language Modeling with Pathways | ✓ Link | 52.9 | PaLM-540B (CoT) | 2022-04-05 | |
| Rethinking embedding coupling in pre-trained language models | ✓ Link | 42.8 | 58.1 | Decoupled | 2020-10-24 |
| PaLM 2 Technical Report | ✓ Link | 73.6 | PaLM 2-L (one-shot) | 2023-05-17 | |
| PaLM 2 Technical Report | ✓ Link | 73.3 | PaLM 2-S (one-shot) | 2023-05-17 | |
| PaLM 2 Technical Report | ✓ Link | 73.3 | PaLM 2-M (one-shot) | 2023-05-17 |