Circuit Hypernetworks for Quantum-Augmented Diffusion Language Models

The study introduces HyperQ, which adds token-conditioned quantum residual branches to a frozen masked-diffusion language model. A lightweight circuit hypernetwork emits token-specific rotation angles, coupling strengths, and measurement axes in a shared sparse structure. An exact classical expression allows evaluation cost to grow linearly with the qubit count, supporting 16 to 64 qubits within a 1.1-billion-parameter backbone. Results show that increasing circuit width raises the average score from 47.65 to 54.30, exceeding classical baselines while fine-tuned on fewer samples.

Across downstream benchmarks, increasing circuit width raises the average score from 47.65 to 54.30.

At 64 qubits, HyperQ exceeds the backbone and its low-rank-adapted counterpart by 4.71 and 3.67 points, respectively.

HyperQ is fine-tuned on 20,000 prompt-response pairs, compared with 200,000 for the classical baselines.