Language Understanding on DCLM evaluation suite (test)
62.9HellaSwag AccuracyLLaMA-3.2
Evaluation results
| Method | Links | ||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LLaMA-3.2Params=1.2B, Tokens=9T2026.05 | 62.9 | 64.8 | 67.8 | 37.6 | 76 | 37 | 75.7 | 62.7 | 79.9 | 60.9 | 65.7 | 36 | 47.9 | 34.4 | 64.3 | 45.8 | 40.5 | 23.9 | 25.3 | 53.1 | |
| LFM-2Params=350M, Tokens=10T2026.05 | 48.3 | 47.3 | 71.5 | 44.5 | 69 | 54.1 | 69.8 | 39.8 | 60.8 | 55.8 | 57.2 | 30.3 | 31.8 | 6.9 | 40.6 | 40.5 | 31.9 | 24.3 | 28.2 | 44.9 | |
| Qwen2Params=500M, Tokens=12T2026.05 | 48 | 48.2 | 59.5 | 31.1 | 67 | 58.2 | 70 | 49.4 | 69.6 | 55.7 | 61.4 | 32.4 | 49.2 | 13.9 | 59.4 | 44.2 | 30.5 | 25.2 | 31.8 | 47.6 | |
| Chronicle Stage 2Params=324M, Tokens=~153B2026.05 | 43.5 | 42.9 | 65.1 | 32.5 | 66 | 19.3 | 69.4 | 38.2 | 66.3 | 51.6 | 56.6 | 21.1 | 29 | 10.1 | 53.8 | 40.5 | 16.7 | 23.9 | 24.8 | 40.6 | |
| Chronicle Stage 1Params=324M, Tokens=~138B2026.05 | 43 | 42.7 | 64.4 | 32.5 | 67 | 23.5 | 68.4 | 39.7 | 64.1 | 50.9 | 57 | 22.1 | 30 | 12.1 | 53.3 | 38.9 | 17.6 | 28.3 | 24.8 | 41.1 | |
| Gemma-3Params=270M, Tokens=6T2026.05 | 40.1 | 39.7 | 58.3 | 28.9 | 67 | 20.7 | 67.6 | 42.9 | 65.2 | 53.6 | 51.7 | 22.3 | 25 | 13 | 54.8 | 43.6 | 21 | 30 | 25.4 | 40.6 | |
| GPT-2Params=124M, Tokens=~10B2026.05 | 31 | 30.8 | 41.7 | 22.4 | 63 | 23 | 62.4 | 32.2 | 57.5 | 50.7 | 54.7 | 13.6 | 5.8 | 0.3 | 28.3 | 42.3 | 9 | 20.9 | 25.8 | 32.4 |