量子糾纏是量子理論的核心基石之一。該現象在低能量尺度下已得到廣泛研究，但其在高能量尺度下的適用性鮮少被討論。本論文聚焦於探討純量希格斯玻色子衰變產生的自旋一 $WW^\ast$ 對的自旋-自旋相關性，特別以大型強子對撞機中超環面儀器的模擬膠子融合機制產生希格斯玻色子為研究對象，分析 $H \to WW^\ast \to \ell \nu \ell \nu$ 衰變頻道中 $WW^\ast$ 對的糾纏性。

首先，本文將介紹量子態斷層掃描，並將其應用於廣義三態最佳化貝爾測試（CGLMP 不等式），以分析糾纏特性。糾纏分析將於真值完備階段進行，並探討不同選擇條件下的結果。為克服模擬真實階段中因微中子難以偵測所帶來的挑戰，本文提出採用具有殘差架構的類神經網路，並在回歸器階段融入物理限制條件，構建從 $WW^\ast$ 分類器到 $W/W^\ast$ 四動量回歸器的完整分析框架。該回歸模型命名為 \underline{P}hysics-\underline{c}onstrained \underline{Res}idual \underline{回歸器}（ PcRes 回歸器），主要應用於擬真階段的分析，並對單一 $W$ 玻色子的預測效能進行評估與呈現。此外，本研究利用 PcRes 回歸器預測之單一 $W$ 玻色子四動量，進一步探討與量子糾纏相關參數的特性。最後，本文提出未來的研究方向，涵蓋若干可增強研究深度與廣度的潛在方法。

Quantum entanglement is the keystone of quantum theory. This phenomenon has been widely studied in the low-energy region; however, the studies that reach the high-energy scale are rarely discussed. In this thesis, we perform a simulation study focusing on the spin-spin correlation of the entangled spin-1 $WW^\ast$ pair from the scalar Higgs boson. We investigate the $H \to WW^\ast \to \ell \nu \ell \nu$ decay channel from the Higgs boson production through the ggF process within the ATLAS detector at the LHC.

First of all, we will introduce the method known as quantum states tomography and apply it to investigate the generalized three-state optimal Bell test, specifically the CGLMP inequality. Our analysis of entanglement will be conducted at the truth-level using various selections up to the preselection criteria. Secondly, to overcome the challenges posed by the hardly detectable neutrinos present in the final states at the reco-level, we leverage the neural network with a residual architecture and incorporate physically informed constraints. A comprehensive discussion from the $WW^\ast$ classifier to the $W/W^\ast$ four-vectors regressor will be provided. The corresponding regressing neural network is called \underline{P}hysics-\underline{c}onstrained \underline{Res}idual \underline{regressor} (PcRes regressor) that will be mainly implemented in the reco-level analysis. The performance of predictions on the individual W bosons will be reported. Furthermore, we will deploy the trained model to investigate the entanglement on the entanglement-relevant parameters using the results predicted by the PcRes regressor. Ultimately, for future directions, this thesis will include some developing approaches that could enhance this research.