Commonsense Reasoning on Winogrande ET (accuracy)
65.1Winogrande ET AccuracyLlama-3.2-3B
Evaluation results
| Method | Links | |
|---|---|---|
| Llama-3.2-3BTraining Status=Continued pretraining, Tokenizer Strategy=default2025.12 | 65.1 | |
| Llama-3.2-3BTraining Status=Continued pretraining, Tokenizer Strategy=prune+ext (16k, naive)2025.12 | 63.2 | |
| Llama-3.2-3BTraining Status=Continued pretraining, Tokenizer Strategy=prune+ext (16k, ours)2025.12 | 63.1 | |
| Llama-3.2-1BTraining Status=Continued pretraining, Tokenizer Strategy=prune+ext (16k, ours)2025.12 | 58.3 | |
| Llama-3.2-1BTraining Status=Continued pretraining, Tokenizer Strategy=default2025.12 | 58.1 | |
| Llama-3.2-1BTraining Status=Continued pretraining, Tokenizer Strategy=prune+ext (16k, naive)2025.12 | 57.4 | |
| Llama-3.2-3BTraining Status=No training, Tokenizer Strategy=default2025.12 | 53 | |
| Llama-3.2-1BTraining Status=No training, Tokenizer Strategy=default2025.12 | 50 |