LARK-Lab/CodeScaler-8B
Full fine-tune vs Skywork/Skywork-Reward-V2-Qwen3-8B. Fine-tune: 92% of the weights changed, across 36 of 36 layers.
Tensors changed: Embeddings 0/1, Norms 106/145, MLP 108/108, Attention 144/144, other 1/1
Full fine-tune vs Skywork/Skywork-Reward-V2-Qwen3-8B. Fine-tune: 92% of the weights changed, across 36 of 36 layers.
Tensors changed: Embeddings 0/1, Norms 106/145, MLP 108/108, Attention 144/144, other 1/1