inference-optimization/Qwen3-Next-80B-A3B-Instruct-GSM8K-MTP-finetuned

Add GSM8K-finetuned MTP head for Qwen3-Next-80B (epoch 0, +22.5% acceptance rate)

initial commit