Oren Data Distillation Experiment Two identical d10 models (100M params) trained to validate the hypothesis that quality-filtered data enables more efficient training. vitalune/nanochat-d10-raw-700m Text Generation • Updated Nov 1, 2025 • 7 vitalune/nanochat-d10-filtered-500m Text Generation • Updated Feb 11 • 5 • 2
Oren Data Distillation Experiment Two identical d10 models (100M params) trained to validate the hypothesis that quality-filtered data enables more efficient training. vitalune/nanochat-d10-raw-700m Text Generation • Updated Nov 1, 2025 • 7 vitalune/nanochat-d10-filtered-500m Text Generation • Updated Feb 11 • 5 • 2