QPredSGG: Hybrid Quantum Predicate Learning for Long-Tailed Scene Graph Generation
Abstract
Scene Graph Generation (SGG) requires relational reasoning over objects and their interactions, but performance is constrained by long-tail predicate imbalance. Classical SGG models rely on dataset statistics, leading to biased predictions toward frequent relations rather than fine-grained semantic predicates. Although debiasing improves mean recall, predicate classification in existing frameworks still relies on high-dimensional classical decision modules with high parameter overhead. This work introduces a resource-efficient hybrid quantum predicate classifier for SGG by replacing the classical predicate head in the CFEN with a QP-Head trained using class-balanced WCE. To the best of our knowledge, this work is among the first evaluations of a hybrid quantum architecture for scene graph predicate classification on Visual Genome 150. We evaluate the impact of qubit count, encoding strategy, entangling topology, and circuit depth. Under a 4-qubit configuration using Amplitude Embedding and Strongly Entangling Layers, the QP-Head projects 4096-dimensional pair features into a 16-dimensional quantum input while using 96 trainable quantum parameters. Operating within this compressed representation space under WCE training, it achieves strong long-tail recognition, reaching a peak mR@100 of 57.25% compared to the 41.1% mR@100 reported in the CFEN baseline under standard training. Protocol-matched ablations indicate that the primary advantage of the quantum head lies in extreme representational compactness rather than an accuracy gain over full-dimensional weighted classical heads: it achieves global recall within 0.19pp of the 2048-dimensional WCE classical baseline while operating from a 128x smaller head feature space. Scaling to 8 qubits maintains robust long-tail performance (55.38% mR@100 with 384 quantum parameters), while depth analysis highlights expressibility-latency trade-offs.
Related
- Learning nuclear mass systematics with hybrid quantum-classical mixtures of experts
- A hybrid quantum-classical neural network for learning to route
- Facial classification Using Hybrid Quantum Machine Learning
- Towards unsupervised representation learning for quantum data: quantum models with inference and generation
- HyQ-LC: Improving Light-Curve Classification with Hybrid Quantum--Classical Learning
BibTeX
@misc{ramkumar2026qpredsgg,
title = {{QPredSGG: Hybrid Quantum Predicate Learning for Long-Tailed Scene Graph Generation}},
author = {Prerana Ramkumar and Nouhaila Innan and Muhammad Shafique},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2606.04689},
eprint = {2606.04689},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2606.04689},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - QPredSGG: Hybrid Quantum Predicate Learning for Long-Tailed Scene Graph Generation
AU - Prerana Ramkumar
AU - Nouhaila Innan
AU - Muhammad Shafique
PY - 2026
DA - 2026/10/24
PB - arXiv
DO - 10.48550/arxiv.2606.04689
UR - https://arxiv.org/abs/2606.04689
AB - Scene Graph Generation (SGG) requires relational reasoning over objects and their interactions, but performance is constrained by long-tail predicate imbalance. Classical SGG models rely on dataset statistics, leading to biased predictions toward frequent relations rather than fine-grained semantic predicates. Although debiasing improves mean recall, predicate classification in existing frameworks still relies on high-dimensional classical decision modules with high parameter overhead. This work introduces a resource-efficient hybrid quantum predicate classifier for SGG by replacing the classical predicate head in the CFEN with a QP-Head trained using class-balanced WCE. To the best of our knowledge, this work is among the first evaluations of a hybrid quantum architecture for scene graph predicate classification on Visual Genome 150. We evaluate the impact of qubit count, encoding strategy, entangling topology, and circuit depth. Under a 4-qubit configuration using Amplitude Embedding and Strongly Entangling Layers, the QP-Head projects 4096-dimensional pair features into a 16-dimensional quantum input while using 96 trainable quantum parameters. Operating within this compressed representation space under WCE training, it achieves strong long-tail recognition, reaching a peak mR@100 of 57.25% compared to the 41.1% mR@100 reported in the CFEN baseline under standard training. Protocol-matched ablations indicate that the primary advantage of the quantum head lies in extreme representational compactness rather than an accuracy gain over full-dimensional weighted classical heads: it achieves global recall within 0.19pp of the 2048-dimensional WCE classical baseline while operating from a 128x smaller head feature space. Scaling to 8 qubits maintains robust long-tail performance (55.38% mR@100 with 384 quantum parameters), while depth analysis highlights expressibility-latency trade-offs.
ER -