kilian-group/KBevo-Qwen3-4B-SFT

Full fine-tune vs Qwen/Qwen3-4B. Fine-tune: 100% of the weights changed, across 36 of 36 layers.

Tensors changed: Norms 137/145, MLP 108/108, Attention 144/144

2 derivatives checked

• kilian-group/KBevo-Qwen3-4B-GRPO: Full fine-tune
• mradermacher/KBevo-Qwen3-4B-SFT-GGUF: Same weights, converted

On the Hub · Downloads on Model Pulse