khazarai/HeisenbergQ-0.5B-RL
TEXT GENERATIONConcurrency Cost:1Model Size:0.5BQuant:BF16Ctx Length:32kPublished:Mar 28, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Warm

HeisenbergQ-0.5B-RL by khazarai is a 0.5 billion parameter model fine-tuned from Qwen2.5-0.5B-Instruct, specifically optimized for quantum physics reasoning. It utilizes GRPO reinforcement learning with custom reward functions to produce structured XML answers, excelling at step-by-step logical reasoning in physics problems. This specialized model is designed for scientific reasoning in math and physics, particularly within the quantum domain.

Loading preview...