Language Modeling on 10-domain bilingual corpus (val)
8.27Validation PerplexityGPT-2
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GPT-2Params=201M, Training=Standard2026.05 | 8.27 | 0 | |
| SNN (AuxCE)Params=194M, Training=AuxCE, Seed=4562026.05 | 8.88 | 89 | |
| SNN (noAuxCE)Params=194M, Training=noAuxCE, Seed=422026.05 | 8.9 | 89 | |
| SNN mean ± std2026.05 | 8.905 | 89 | |
| SNN (AuxCE)Params=194M, Training=AuxCE, Seed=1232026.05 | 8.91 | 89 | |
| SNN (noAuxCE)Params=194M, Training=noAuxCE, Seed=1232026.05 | 8.93 | 89 | |
| GPT-2Params=124M, Training=Standard2026.05 | 8.96 | 0 |