Text Generation on OpenWebText (OWT) (test)
0.85MAUVEAR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ARArchitecture=12-layer Llama-2-style2025.06 | 0.85 | 15.05 | 25.24 | |
| AR + DiffusionBlocksArchitecture=12-layer Llama-2-style, Blocks=B=4, Memory Reduction=4x2025.06 | 0.82 | 14.99 | 26.33 |