G4-31B-stock45-GRPO-r4b100
SLERP of gemma-4-31B-it (0.45) and G4-31B-SFT-v10-2 (0.55), with the GRPO
diversity LoRA (run 4b step 100, trained strength) merged in. The
SFT-leaning sibling of G4-31B-stock55-GRPO-r4b100. Same chat template and
usage as gemma-4-31B-it.
