espressovi
/

BODHI-gemma-3-12b-distil

Text Generation

Model card Files Files and versions

Artifact for the BODHI project. This math model is created by Long-CoT distillation starting from google/gemma-3-12b-pt.

Performance

The following results are at T=1.0, 16K tokens and pass@8.

AIME25:

Downloads last month: 62

Safetensors

Model size

12B params

Tensor type

BF16

·

Model tree for espressovi/BODHI-gemma-3-12b-distil

Base model

google/gemma-3-12b-pt

Finetuned

(113)

this model

Finetunes

Dataset used to train espressovi/BODHI-gemma-3-12b-distil

Collection including espressovi/BODHI-gemma-3-12b-distil

BODHI

Artifacts for the paper titled "BODHI: Can LLMs Branch Out and Discover Heterogeneous Inferences?" • 18 items • Updated Jul 17