---
title: "Staged Hybridisation for Visual Quantum Reinforcement Learning via Knowledge Distillation"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2606.30520v1"
permalink: "https://feed.astralor.com/items/itm_cfc5c1c0f4e525e26ef4d541"
published: "2026-06-29T16:28:57.000Z"
score: 56
channel: "ai"
tags: ["Variational Quantum Circuit","Knowledge Distillation","Visual Quantum Reinforcement Learning","Quantum Machine Learning"]
---

# Staged Hybridisation for Visual Quantum Reinforcement Learning via Knowledge Distillation

## At a glance

This study proposes a staged knowledge distillation strategy to address the challenges of visual quantum reinforcement learning (QRL). A classical visual teacher is first trained, its encoder frozen as a feature interface, and the teacher's policy behavior is distilled into compact downstream heads (classical or VQC-based). Experiments on CartPole Pixels and Acrobot Pixels show that shallow VQC heads can acquire non-trivial visual control behavior, angle-encoded VQC heads achieve near-teacher performance, while amplitude-encoded heads are extremely compact but more fragile. This approach reframes visual QRL as a compact-head learning problem, providing a practical route for training small quantum-compatible policies outside the standard end-to-end loop.

## Key points

- Staged knowledge distillation avoids the difficulty of end-to-end training visual quantum networks by first training a classical teacher and then distilling the policy to compact downstream heads.
- The frozen teacher encoder serves as a feature interface, allowing VQC-based students to be fairly compared with classical controls under the same representation.
- Angle-encoded VQC heads achieve near-teacher performance while remaining compact, whereas amplitude-encoded heads are extremely compact but sensitive to budget and simulation time.

## Full text

Visual environments are a demanding setting for quantum reinforcement learning (QRL): high-dimensional observations, unstable RL optimisation, and constrained variational quantum circuits (VQCs) are difficult to train jointly. This paper studies knowledge distillation (KD) as a staged hybridisation strategy for visual QRL. Instead of training a hybrid visual agent end-to-end from pixels, we first train a classical visual teacher, freeze its encoder as a feature interface, and distil the teacher's policy behaviour into compact downstream heads. These heads can be classical or VQC-based, enabling small quantum-compatible students to be evaluated under the same frozen representation as compact classical controls. We evaluate the pipeline on CartPole Pixels and Acrobot Pixels. The results show that staged KD enables shallow VQC heads to acquire non-trivial visual-control behaviour in settings where direct pixel-based training would be substantially more difficult. Angle-encoded VQC heads retain near-teacher performance, while amplitude-encoded heads push compactness to an extreme regime, at the cost of greater fragility, stronger budget sensitivity, and higher simulation time. Overall, staged KD reframes visual QRL as a compact-head learning problem, opening a practical route for training small quantum-compatible policies outside the standard end-to-end RL loop.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_cfc5c1c0f4e525e26ef4d541