Update focused on increasing reasoning density and reducing unneeded verbosity via intermittently rewarding compactness on GSM8K with GRPO. Hopefully generalizes as she wanted it to. Creative writing skill seems intact.
- Downloads last month
- 20
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for Lambent/Iris-12B-v1.2
Base model
google/gemma-4-12B Finetuned
google/gemma-4-12B-it Finetuned
Lambent/Iris-12B-gemma-4-it-qat Finetuned
Lambent/Iris-12B-v1.1