HexBench Qwen3-Coder-30B-A3B adapter (dpo-b010)

SFT(ep3)-merged + DPO beta=0.1 (genuine>-scalar pairs)

NOTE: continued-training adapter; the standalone merged best model is hitmanonholiday/qwen3coder30b-hexbench-best.

Downloads last month
26
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for hitmanonholiday/qwen3coder30b-hexbench-dpo-b010