Quantum Verifiable Rewards for Post-Training Qiskit Code Assistant
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908508635529216 |
|---|---|
| author | Dupuis, Nicolas Tiwari, Adarsh Mroueh, Youssef Kremer, David Faro, Ismael Cruz-Benito, Juan |
| author_facet | Dupuis, Nicolas Tiwari, Adarsh Mroueh, Youssef Kremer, David Faro, Ismael Cruz-Benito, Juan |
| contents | Qiskit is an open-source quantum computing framework that allows users to design, simulate, and run quantum circuits on real quantum hardware. We explore post-training techniques for LLMs to assist in writing Qiskit code. We introduce quantum verification as an effective method for ensuring code quality and executability on quantum hardware. To support this, we developed a synthetic data pipeline that generates quantum problem-unit test pairs and used it to create preference data for aligning LLMs with DPO. Additionally, we trained models using GRPO, leveraging quantum-verifiable rewards provided by the quantum hardware. Our best-performing model, combining DPO and GRPO, surpasses the strongest open-source baselines on the challenging Qiskit-HumanEval-hard benchmark. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_20907 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Quantum Verifiable Rewards for Post-Training Qiskit Code Assistant Dupuis, Nicolas Tiwari, Adarsh Mroueh, Youssef Kremer, David Faro, Ismael Cruz-Benito, Juan Quantum Physics Artificial Intelligence Qiskit is an open-source quantum computing framework that allows users to design, simulate, and run quantum circuits on real quantum hardware. We explore post-training techniques for LLMs to assist in writing Qiskit code. We introduce quantum verification as an effective method for ensuring code quality and executability on quantum hardware. To support this, we developed a synthetic data pipeline that generates quantum problem-unit test pairs and used it to create preference data for aligning LLMs with DPO. Additionally, we trained models using GRPO, leveraging quantum-verifiable rewards provided by the quantum hardware. Our best-performing model, combining DPO and GRPO, surpasses the strongest open-source baselines on the challenging Qiskit-HumanEval-hard benchmark. |
| title | Quantum Verifiable Rewards for Post-Training Qiskit Code Assistant |
| topic | Quantum Physics Artificial Intelligence |
| url | https://arxiv.org/abs/2508.20907 |