Benyucong/sft_quantum_circuit_gen_4B

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 26, 2025License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

Benyucong/sft_quantum_circuit_gen_4B is a 4 billion parameter instruction-tuned language model, fine-tuned from Qwen/Qwen3-4B-Instruct-2507. Developed by Benyucong, this model specializes in generating and optimizing quantum circuits, particularly for tasks like solving vertex cover problems. It is designed to produce valid QASM 3.0 code based on natural language prompts, making it suitable for quantum computing research and development.

Loading preview...

Model Overview

Benyucong/sft_quantum_circuit_gen_4B is a 4 billion parameter language model specifically fine-tuned for quantum circuit generation and optimization. It is based on the Qwen/Qwen3-4B-Instruct-2507 architecture and leverages the research presented in "Agent-Q: Fine-Tuning Large Language Models for Quantum Circuit Generation and Optimization" (arXiv:2504.11109). The model was trained using the TRL framework, focusing on supervised fine-tuning (SFT).

Key Capabilities

  • Quantum Circuit Generation: Generates quantum circuits in QASM 3.0 format from natural language descriptions.
  • Optimization Focus: Aims to provide optimal parameters within the generated circuits.
  • Problem Solving: Demonstrated capability for tasks such as solving the vertex cover problem on given graph structures.
  • Instruction Following: Responds to detailed user prompts regarding qubit count, layers, and specific problem parameters.

Use Cases

This model is particularly well-suited for:

  • Quantum Computing Research: Assisting researchers in quickly prototyping and exploring quantum circuit designs.
  • Educational Tools: Providing examples and solutions for quantum algorithm design.
  • Automated Circuit Design: Generating initial circuit structures for further refinement or simulation.

Training Details

The model underwent supervised fine-tuning (SFT) using TRL (version 0.15.2), with Transformers 4.55.0 and Pytorch 2.7.1. Further details on the training process and performance can be explored via the Weights & Biases run linked in the original repository.

Finetunes
1 models