Language Modeling on LAMBADA (CE, Accuracy)
0.5823Cross-Entropy (CE)SYNPRO
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SYNPROModel Scale=1.1B, Organic Tokens=2.2B2026.05 | 0.5823 | 40.07 | |
| ReProModel Scale=1.1B, Organic Tokens=2.2B2026.05 | 0.6136 | 37.71 | |
| RepeatModel Scale=1.1B, Organic Tokens=2.2B2026.05 | 0.6645 | 35.77 | |
| SYNPROModel Scale=400M, Organic Tokens=0.8B2026.05 | 0.7907 | 29.52 | |
| ReProModel Scale=400M, Organic Tokens=0.8B2026.05 | 0.8175 | 27.42 | |
| RepeatModel Scale=400M, Organic Tokens=0.8B2026.05 | 0.8714 | 25.62 |