minpeter/calculator-agent-qwen3-0.6b

It's trash. It has a serious drop in score compared to the base model. But it was my first RL learning and it was monumental.

eval

minpeter/calculator-agent-qwen3-0.6b: Accuracy: 15.19% (24/158) minpeter/Qwen3-0.6B-Instruct: Accuracy: 27.22% (43/158)