ermiaazarkhalili/LFM2-700M-GRPO-NuminaMath-10K
Adapter vs LiquidAI/LFM2-700M. ermiaazarkhalili/LFM2-700M-GRPO-NuminaMath-10K is a LoRA adapter, not full weights: rank 16, on k_proj, q_proj, v_proj in 6 layers (18 adapted matrices). It modifies LiquidAI/LFM2-700M at load time.