Knowledge-focused evaluation on MixEval Hard
21.8AccuracyFineInstructions
Evaluation results
| Method | Links | |
|---|---|---|
| FineInstructionsPre-training Corpus=Nemotron-CC Corpus (300B), Model Parameters=1.8B2026.01 | 21.8 | |
| FineInstructionsPre-training Corpus=IPT Corpus (23B), Model Parameters=1.8B2026.01 | 19.2 | |
| Q&APre-training Corpus=Nemotron-CC Corpus (300B), Model Parameters=1.8B2026.01 | 18.9 | |
| WRAPPre-training Corpus=Nemotron-CC Corpus (300B), Model Parameters=1.8B2026.01 | 18.4 | |
| Standard Pre-TrainingPre-training Corpus=Nemotron-CC Corpus (300B), Model Parameters=1.8B2026.01 | 17.1 | |
| IPTPre-training Corpus=IPT Corpus (23B), Model Parameters=1.8B2026.01 | 16.7 | |
| Nemotron-CCPre-training Corpus=Nemotron-CC Corpus (300B), Model Parameters=1.8B2026.01 | 16.7 | |
| Standard Pre-TrainingPre-training Corpus=IPT Corpus (23B), Model Parameters=1.8B2026.01 | 14 |