Robust Quantum Algorithmic Binary Decision-Making on Gaussian Signals

Extracted source text · arXiv 2601.16081

\documentclass[conference]{IEEEtran}
\IEEEoverridecommandlockouts

\usepackage[
  top=0.73in,
  bottom=1.0in,
  left=0.625in,
  right=0.625in
]{geometry}

% \usepackage{cite}

\usepackage{balance}
\usepackage{cmap}
\usepackage[utf8]{inputenc}
\usepackage[T1]{fontenc}
\usepackage{graphicx}

% \usepackage{setspace}
\usepackage{multirow, makecell}

%Math and Physics Packages
\usepackage{amsmath}
\allowdisplaybreaks %allow page break in multiline eqns (amsmath command)
\usepackage{amssymb}
\usepackage{amstext}
\usepackage{amsfonts}
\usepackage{amsthm}
\usepackage{mathtools}
\DeclarePairedDelimiter\ceil{\lceil}{\rceil}
\DeclarePairedDelimiter\floor{\lfloor}{\rfloor}
\usepackage{bm}
\usepackage{braket}
\usepackage{physics}
\usepackage{enumitem}

\usepackage[caption=false]{subfig}
\usepackage{overpic}


\newcommand{\YL}[1]{{\color{red}{YL: #1}}}
\newcommand{\AM}[1]{{\color{magenta}{AM: #1}}}


\usepackage{hyperref}
\hypersetup{colorlinks,linkcolor=blue,anchorcolor=blue,citecolor=blue,urlcolor=blue}

\usepackage{dcolumn}% Align table columns on decimal point
\usepackage{bm}
\usepackage{array}
\usepackage{xcolor}
\usepackage{float}
\usepackage{etaremune}
\usepackage{enumitem}
\usepackage{tablefootnote}
\usepackage{url}
% \usepackage{stfloats}
% \usepackage{makecell,boldline}
% \usepackage{tablefootnote}
% \usepackage{cleveref}
\usepackage{tikz}
\usepackage{qcircuit}
\usepackage{subfig}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\usepackage{subfiles}
\usepackage{etoolbox}

\theoremstyle{plain}
\newtheorem{thm}{Theorem}
\newtheorem{cor}[thm]{Corollary}
\newtheorem{lem}[thm]{Lemma}
\newtheorem{prop}[thm]{Proposition}
\theoremstyle{definition}
\newtheorem{defn}[thm]{Definition}
\theoremstyle{remark}
\newtheorem*{remark}{Remark}

\makeatother

%%%%%%%%%%
%%%%%%%%%%

\usepackage{orcidlink}
% \usepackage{authblk}

%%%%%%%%%% Define tikz commands for a short picture of Heisenberg plaquette (only needed if you don't upload figure PDFs)


\begin{document}

\title{Robust Quantum Algorithmic Binary Decision-Making on Displacement Signals

% \\
% \thanks{This work is supported by the U.S. Department of Energy, Office of Science, Advanced Scientific Computing Research, under contract number DE-SC0025384.}
}

%author list format 1
% \author[1]{Aishwarya Majumdar\orcidlink{0009-0008-2800-0455},}
% \author[1,2,3]{Yuan Liu\orcidlink{0000-0003-1468-942X}}
% \affil[1]{\textit{Department of Electrical and Computer Engineering, North Carolina State University, Raleigh,
% NC 27606, USA}}
% \affil[2]{\textit{Department of Computer Science, North Carolina State University, Raleigh,
% NC 27606, USA}}
% \affil[3]{\textit{Department of Physics and Astronomy, North Carolina State University, Raleigh,
% NC 27695, USA}
% }


% author list format 2, disable \usepackage{authblk}
\author{\IEEEauthorblockN{ Aishwarya Majumdar
\orcidlink{0009-0008-2800-0455}
}
\IEEEauthorblockA{\textit{Department of Electrical and Computer Engineering, } \\
\textit{North Carolina State University, Raleigh, NC 27606, USA}\\
amajumd4@ncsu.edu}

\and

\IEEEauthorblockN{Yuan Liu
\orcidlink{0000-0003-1468-942X}
}
\IEEEauthorblockA{\textit{Department of Electrical and Computer Engineering,}
\\
\textit{Department of Computer Science,}
\\
\textit{Department of Physics and Astronomy,}
\\
\textit{North Carolina State University, Raleigh, NC 27606, USA}
\\
q\_yuanliu@ncsu.edu}
}


\maketitle

\begin{abstract}
A relevant signal in the quantum domain may manifest as a displacement or a phase shift operator in the bosonic phase space. For a real parameter \(\beta\) embedded in such a displacement operator, the task of determining if \(\beta \in [\beta_{-th}, \beta_{+th}]\) for real asymmetric thresholds (\(\beta_{-th} \ne -\beta_{+th}\)) is a binary decision problem. We propose a framework based on \emph{generalized quantum signal processing interferometry} (GQSPI) on hybrid qubit-bosonic oscillator systems that addresses this parameter detection problem by recasting the practical task of active binary hypothesis testing on quantum systems to that of a polynomial approximation. We achieve a small decision error probability \(p_{err}\) on the order of \(\mathcal{O}(\frac{1}{d}\log{(d)})\), with \(d\) as the circuit depth. We analyze the protocol when (i) \(\beta\) is a deterministic parameter, and (ii) when \(\beta\) is drawn randomly from a known prior distribution. The performance of the sensing protocol under dephasing noise is also shown to be robust. We further extend our protocol from two thresholds to more general multi-threshold cases as well. Overall, the proposed framework enables decision-making over arbitrary thresholds for any general displacement signal in a single or a few shots.
\end{abstract}
% \vspace*{-0.5\baselineskip}


\section{Introduction} 
\label{sec:intro}

Binary decision-making lies at the heart of several fundamental applications of information science, including sensing, communication, and computing. 
Classical detection theory \cite{wald1950statistical, van1971detection, 1988svPoor} offers a formal approach to decision problems via hypothesis testing techniques. 
This includes optimal detector designs such as the Neyman-Pearson \cite{10.1098/rsta.1933.0009, 9500a98b-1cec-3671-80dd-bc45a0de49f0} detectors that take into account cost functions and receiver operating characteristics, and the Bayesian detectors \cite{1057460, doi:10.1137/0103017, doi:10.1137/0104006} that minimize the probability of a decision error based on known information. Similarly, Helstrom and Holevo developed the early theory for quantum detection and hypothesis testing \cite{1969JSP1231H, 1968IJTP137H, HOLEVO1973337, doi:10.1137/1123048}, providing the foundations for designing optimal measurement operators and the minimum error bounds involved in discriminating quantum states, given prior probabilities. 

Practical applications of quantum detection theory include quantum sensing methods \cite{RevModPhys.89.035002, PhysRevLett.96.010401} that often deal with detection of rare astrophysical events such as dark matter interactions \cite{Bass2024,PhysRevLett.133.021801, ma2025qubitsmultilevelquantumsensing, cwx5-2n1y,
rv43-54zq} or gravitational waves \cite{ligo2015squeezing, bothwell2022resolving, ligo2016observation}, where the presence of minute signals must be distinguished from noise. For quantum communication and computing \cite{von1955mathematical, 10.5555/3240076, wilde2013quantum, nielsen2000quantum}, the task of determining an encoded message transmitted in presence of noise is central to the design of optimal decoders and to achieving the channel capacity \cite{PhysRevLett.66.1119, PhysRevA.54.1869, PhysRevA.51.2738}. 

Properties unique to quantum systems such as entanglement and superposition have enabled the performance of parameter estimation techniques to go beyond the classical limits, also known as the \emph{standard quantum limit} (SQL), and achieve Heisenberg-limited (HL) scaling. Likewise, for binary detection problems, quantum protocols, such as quantum illumination, \cite{doi:10.1126/science.1160627,PhysRevLett.101.253601} have shown improvements over classical methods under identical constraints. However, decision problems in the quantum domain are far more delicate because quantum signals are often buried under classical environmental noise, and quantum states are prone to decoherence, rapidly decaying into other states during the detection process. Although several techniques have been proposed to overcome these hurdles \cite{PhysRevLett.130.070802, Kuniyil_2022, Myatt2000, PhysRevB.77.174509}, a systematic approach offering provable improvements in decision accuracy for active detection of quantum displacement signals is largely missing. 

 
 
In this work, we establish a framework for solving detection problems in which an underlying displacement signal parameter may be deterministic or a random variable with a known prior distribution, with the goal of determining whether the parameter lies between two given real thresholds (say, $\beta_{-th}$, and $\beta_{+th}$). The framework incorporates non-Gaussian resources by applying \emph{generalized quantum signal processing} (GQSP) algorithm \cite{PRXQuantum.5.020368} on hybrid qubit-bosonic oscillator systems, utilizing the infinite-dimensional Hilbert space of the oscillator to sense continuous-variable displacement signal parameter and entangling it with the discrete qubit measurement outcome as the binary decision result. Our approach, \emph{generalized quantum signal processing interferometry} (GQSPI) improves upon the \emph{quantum signal processing interferometry} (QSPI) as proposed in \cite{SinananSingh2024singleshotquantum}
by going beyond symmetric thresholding problems ($\beta_{-th} = -\beta_{+th}$) arising from limited expressivity of the \emph{quantum signal processing} (QSP) algorithm \cite{low2016methodology, low2019hamiltonian, PhysRevLett.118.010501}.
Additionally, we show that the GQSPI protocol is essentially robust against dephasing noise on the oscillators.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%


\section{Background} \label{sec:bg}
\subsection{Notations} \label{sec:bg-notations}

Throughout the paper, we refer to the canonical position and momentum operators of harmonic oscillators as $\hat{x}$ and $\hat{p}$, satisfying $[\hat{x}, \hat{p}] = i$, assuming $\hbar = 1$, mass $m=1$, and frequency $\omega=1$. $a$ and $a^{\dagger}$ are the bosonic annihilation and creation operators satisfying $[a, a^{\dagger}] = 1$ with $\hat{n} = a^{\dagger}a$ being the photon number operator. Using `Standard units' \cite{4rf7-9tfx}, these operators are related as $a = \frac{1}{\sqrt{2}}(\hat{x} + i\hat{p}), a^{\dagger} = \frac{1}{\sqrt{2}}(\hat{x} - i\hat{p})$ and $\hat{x} = \frac{1}{\sqrt{2}}(a + a^{\dagger}), \hat{p} =-i\frac{1}{\sqrt{2}}( a - a^{\dagger})$. 
The vacuum or ground state of a harmonic oscillator is denoted by $\ket{0}_{osc}$ which is a Gaussian state. 
$I_{osc}$ represents the identity operation on the oscillator. 

% We denote the qubit ground state by $\ket{\downarrow}$ and the excited state by $\ket{\uparrow}$. $\sigma_x, \sigma_y, \sigma_z$ are the qubit Pauli X, Y and Z operators, respectively.
We denote the qubit ground state and the excited state by $\ket{\downarrow}$ and $\ket{\uparrow}$, respectively. $\sigma_x, \sigma_y, \sigma_z$ are the qubit Pauli operators.

\subsection{Tools} \label{sec:bg-tools}
Of the several advances in quantum algorithms in the past decade, Quantum Signal Processing (QSP) \cite{low2016methodology, low2019hamiltonian, PhysRevLett.118.010501} and its generalized form, GQSP \cite{PRXQuantum.5.020368} have proven to be exceptionally versatile algorithms in terms of their applicability to solutions for a wide variety of quantum problems, including hamiltonian simulation, quantum search, and factoring \cite{MRTC21}. The standard form of QSP involves interleaving two types of qubit rotation operators repeatedly, namely a signal operator encoding a parameter of interest, and a signal processing operator. The overall resultant unitary at the end of this protocol realizes a polynomial transformation on the parameter. This process of interleaving two different families of quantum operators forms the backbone of the binary decision approach presented in the following section.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\section{Main Problem and Key Results}
We outline the main problem in Sec.~\ref{sec:mp_ta-problem}, followed by the technical approach, and key results of this work in the subsequent subsections. We first introduce GQSP for hybrid qubit–oscillator systems (Theorem~\ref{thm-1}) in Sec.~\ref{sec:mp_ta-hybrid-gqsp}, which forms the foundation for the GQSPI protocol presented in Sec.~\ref{sec:mp_ta-gqspi-thm} (Theorem~\ref{thm-2}). We analyze error scaling in Sec.~\ref{sec:results-error}, discuss the stochastic setting in Sec.~\ref{sec:mp_ta-beta-rv}, and examine the robustness of the protocol under dephasing noise in Sec.~\ref{sec:mp_ta-dephasing}. Finally, we present simulation results in Sec.~\ref{sec:results}. 

\subsection{Main Problem} \label{sec:mp_ta-problem}

The signal is assumed to interact with the qubit-bosonic oscillator system such that it effectively applies a position change of $\beta$ to the oscillator state via the operator $S_\beta$. Such an operator, also known as displacement signal, in the joint qubit-oscillator space is given by,
\begin{align}
    S_\beta = I \otimes e^{i\beta \hat{p}} =
    \begin{bmatrix}
        e^{i \beta \hat{p}}  & 0 \\
        0 & e^{i \beta \hat{p}}
    \end{bmatrix}, \quad \beta \in \mathbb{R}
    \label{s-beta}.
\end{align}

With the goal to determine whether the value of $\beta$ lies within the range $[\beta_{-th}, \beta_{+th}]$,
the GQSPI protocol prepares the qubit–oscillator system so as to minimize the overall binary decision error. The main problem we solve in this work is:

\emph{Given thresholds $\beta_{-th}$,  $\beta_{+th} \in \mathbb{R}$, determine whether the parameter $\beta \in \mathbb{R}$, embedded in a displacement operator $S_{\beta}$ 
in the qubit-oscillator space, 
lies within the range $[\beta_{-th}, \beta_{+th}]$ or outside of it, with a small decision error probability $p_{err}$, using a quantum circuit constructed using qubit rotation gates and qubit-oscillator entangling gates.}


We assume that the qubit is in ground state $\ket{\downarrow}$, and the oscillator is in the vacuum state $\ket{0}_{osc}$ at the beginning of the protocol. Following the GQSPI protocol, the qubit is measured along the Pauli-Z basis, where the measurement outcome $\downarrow$ signifies that the parameter $\beta$ is in the range $[\beta_{-th}, \beta_{+th}]$, while measuring $\uparrow$ implies that $\beta$ does not belong to this range. We denote $P(M = \downarrow| \beta)$ as the probability of measuring $\downarrow$ on the qubit for a given value of $\beta$.

\subsection{ GQSP Theorem for Qubit-Oscillator Systems} \label{sec:mp_ta-hybrid-gqsp}


Consider the following single-qubit rotation parametrized by angles $\{\theta, \phi, \lambda\}$
\begin{align}
    R(\theta, \phi, \lambda) 
    &= \label{eq:r}
    \begin{bmatrix} 
        e^{i(\lambda+ \phi)}\cos{\theta} & e^{i\phi}\sin{\theta} \\
        e^{i\lambda}\sin{\theta} & 
        -\cos{\theta}
    \end{bmatrix} 
\end{align}
and an entangling operator between the qubit and the oscillator
\begin{align*}
    W_z(\hat{\omega}) = \begin{bmatrix}
        \hat{\omega}(\hat{x}, \hat{p}) & 0 \\
        0 & \hat{\omega}^{-1}(\hat{x}, \hat{p})
    \end{bmatrix},
\end{align*}
which applies operator $\hat{\omega}(\hat{x}, \hat{p})$ to the oscillator state if the qubit state is $\ket{\downarrow}$ and $\hat{\omega}^{-1}(\hat{x}, \hat{p})$ if the qubit state is $\ket{\uparrow}$. As a building block, we define \emph{generalized quantum signal processing} (GQSP) for hybrid qubit-oscillator systems as follows:
\begin{thm}\label{thm-1}
    A quantum circuit on hybrid qubit-oscillator system parameterized by the set of angles $\vec{\theta} = \{\theta_0, \theta_1 \cdots \theta_d\}, ~\vec{\phi} = \{\phi_0, \phi_1 \cdots \phi_d\}$, and $\lambda_0$ realizes a block-encoding of a degree-$d$ complex Laurent polynomial transformation $P(\hat{\omega})$ on an oscillator operator $\hat{\omega}(\hat{x}, \hat{p})$ using the following sequence given by $G_{\vec{\theta}, \vec{\phi}, \lambda_0}(\hat{\omega})$:
\begin{align}
    &\nonumber
    G_{\vec{\theta}, \vec{\phi}, \lambda_0}(\hat{\omega}) 
    \\ 
    \nonumber
    &= \left(\prod_{i=1}^{d} (R(\theta_i, \phi_i, 0)\otimes I_{osc})W_z(\hat{\omega})\right)(R(\theta_0, \phi_0, \lambda_0)\otimes I_{osc}) ,
    \\ 
    &= \begin{bmatrix}
        P_d(\hat{\omega}) &  -Q^{\dagger}_d(\hat{\omega})
        \\
        Q_d(\hat{\omega}) 
        & P^{\dagger}_d(\hat{\omega})
    \end{bmatrix},
\end{align}
such that,
\begin{itemize}
    \item $P_d(\hat{\omega})$ and $Q_d(\hat{\omega})$ are both degree-$d$ polynomials in $\hat{\omega}$, 
    \item $|P_d(\hat{\omega})|^2 + |Q_d(\hat{\omega})|^2 = 1 ~\forall~\hat{\omega}$, 
    \item $P_d(\hat{\omega})$ and $Q_d(\hat{\omega})$ are of the form,
    \\
    $P_d(\hat{\omega}) = \sum_{n=-d}^{d} p_n \hat{\omega}^n,
    \quad
    Q_d(\hat{\omega}) = \sum_{n=-d}^{d} q_n \hat{\omega}^n$ where
    $n = \{-d, -d+2, \cdots, d-2, d\}$, and $p_n, q_n \in \mathbb{C}$.
\end{itemize}
\end{thm}
Similar form of GQSP for hybrid qubit-bosonic oscillator systems was also independently obtained by \cite{hong2025oscillatorqubitgeneralizedquantumsignal}.
The nature of polynomials and relationships between polynomial coefficients can be found in Appendix~\ref{sec:appndx-GQSP-Poly-Forms}.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{ Generalized Quantum Signal Processing Interferometry} \label{sec:mp_ta-gqspi-thm}

\begin{figure}[ht]
    \centering
    \includegraphics[width=0.95\linewidth]{diagrams/Paper-GQSPI.drawio.pdf}
    \vspace*{-1.2\baselineskip}
    \caption{Quantum circuit for Generalized Quantum Signal Processing Interferometry (GQSPI). Note that for the qubit rotation gate in Eq. \eqref{eq:r}, $R^{\dagger}(\theta, \phi, \lambda) = R(\theta, -\lambda,  -\phi)$.}
    \label{fig:gqspi-setup}
    \vspace*{-1\baselineskip}
\end{figure}


Let the qubit-oscillator entangling operator $W_z(\hat{\omega})$ be a conditional displacement gate on the oscillator, conditioned on the qubit state, given by, 
\begin{align}
    \mathcal{D}_c(\alpha) = e^{(\alpha a^{\dagger} - \alpha^* a)\sigma_z} 
\end{align}
where $a$ and $a^{\dagger}$ are the bosonic annihilation and creation operators on the oscillator as noted in Sec. \ref{sec:bg-notations}.
In particular, for $\alpha = i \kappa / \sqrt{2}$ where $\kappa \in \mathbb{R}$, the $\mathcal{D}_c(\alpha)$ operator imparts a momentum kick to the oscillator by $+\kappa$ if the qubit state is $\ket{\downarrow}$ and $-\kappa$ if the qubit state is $\ket{\uparrow}$,
\begin{align}
    \mathcal{D}_c(i \kappa / \sqrt{2}) = e^{i \kappa \hat{x} \hat{\sigma}_z} = \begin{bmatrix}
        e^{i \kappa \hat{x}} & 0 
        \\
        0 & e^{-i \kappa \hat{x}}
    \end{bmatrix}
    \label{Dc-def}.
\end{align}
It can be observed that the $\mathcal{D}_c(i \kappa / \sqrt{2})$ gate is periodic in the position eigen state $\ket{x}_{osc}$ with periodicity $T_{\mathcal{D}} = \frac{2\pi}{\kappa}$ in the bosonic phase-space.


For the interferometry protocol, we sandwich the displacement signal containing operator $S_{\beta}$ between a degree-$d$ GQSP sequence and its inverse sequence, 
\begin{align}
     U(\beta, \kappa)
     &= \label{eq:U-def}
     G^{-1}_{d,\vec{\theta}, \vec{\phi}, \lambda_0}(\hat{x}) ~S_{\beta}~ G_{d,\vec{\theta}, \vec{\phi}, \lambda_0}(\hat{x})
\end{align}
where the GQSP sequence consists of qubit rotation gates $R(\theta, \phi, \lambda)$ and qubit-oscillator entangling gate $\mathcal{D}_c(i \kappa / \sqrt{2})$ as follows:
\begin{align}
     G_{d,\vec{\theta}, \vec{\phi}, \lambda_0}(\hat{x}) 
     &= \nonumber
     \left[(R(\theta_d, \phi_d, 0)\otimes I_{osc})D_c(i\kappa/\sqrt{2})\right] 
     \\
     & \nonumber
     \cdots \times\left[(R(\theta_1, \phi_1, 0)\otimes I_{osc})D_c(i\kappa/\sqrt{2})\right] 
     \\
     & \label{eq:gsqp-expanded}
     \times
     (R(\theta_0, \phi_0, \lambda_0)\otimes I_{osc}) 
\end{align}


From the hybrid GQSP Theorem \ref{thm-1}, $G_{\vec{\theta}, \vec{\phi}, \lambda_0}(\hat{x})$ can be expressed as, 
\begin{align*}
    G_{\vec{\theta}, \vec{\phi}, \lambda_0}(\hat{x}) 
    &=
    \begin{bmatrix}
         P_{d}(\hat{x}) &  -Q^{\dagger}_{d}(\hat{x})
         \\
         Q_{d}(\hat{x}) &  P^{\dagger}_{d}(\hat{x})
    \end{bmatrix}
\end{align*}
where, $P_{d}(\hat{x}) = \sum_{n=-d}^{d} p_n e^{i n\kappa \hat{x}} $ and $Q_{d}(\hat{x}) = \sum_{n=-d}^{d} q_n e^{i n\kappa \hat{x}}$ with $p_n, q_n \in \mathbb{C}$.

By choosing appropriate phase angles $\vec{\theta} = \{\theta_0, \theta_1 \cdots \theta_d\}, ~\vec{\phi} = \{\phi_0, \phi_1 \cdots \phi_d\}$, and $\lambda_0$, the probability of measuring the qubit in state $\ket{\downarrow}$ for given $\beta$, denoted by $P(M = \downarrow| \beta)$ can be made close to 1 for $\beta \in [\beta_{-th}, \beta_{+th}]$ and close to 0 when $\beta$ is outside this range. 

The probability of measuring $\ket{\downarrow}$ for a given $\beta$ is obtained to be,
\begin{align}
    P(M = \downarrow| \beta) 
    &= \nonumber
   \langle 0|_{osc} U^{\dagger}(\beta, \kappa)_{00}U(\beta, \kappa)_{00}|0 \rangle _{osc}
   \\
   &
   = \label{eq:csbeta}
   \sum_{s = -d}^{d} c_s e^{i(2\kappa) \beta s}
   \end{align} 
   \begin{align}
   &\text{with,}\quad 
    c_s 
    = \nonumber
      \sum_{n, m, r = -d}^{d}
    \left( p_n p^{*}_{m} +  q_n q^{*}_{m} \right)
    \\
    & 
    \label{eq:qspi_eq_20}
    \times
    \left( p^{*}_{n + 2s} p_{m + 2r} +  q^{*}_{n + 2s} q_{m + 2r} \right) 
     e^{- \kappa^2 (r - s)^2}
     \\ \nonumber
     &\text{where } s,r,m,n = \{-d, -d+2, \cdots, d-2, d\}.
\end{align}
$U(\beta, \kappa)_{00}$ in Eq. \eqref{eq:csbeta} denotes the top-left element of the operator $U(\beta, \kappa)$ which can be expressed as,
\begin{align}
    U(\beta, \kappa)_{00} 
    &= \label{eq:U00}
    \sum_{n,m=-d}^{d} \left( p^{*}_n p_m +  q^{*}_n q_m \right)e^{-i n\kappa \hat{x}} e^{i\beta \hat{p}} e^{i m \kappa \hat{x}}
\end{align}

It is interesting to note that $P(M = \downarrow| \beta)$ is a polynomial in $e^{i(2\kappa) \beta}$, and has a periodicity of $T_P = \frac{\pi}{\kappa}$. Therefore, it is necessary to choose $\kappa$ appropriately such that $\beta \in \left(-\frac{\pi}{2\kappa}, \frac{\pi}{2\kappa}\right)$ so as to cover the entire range of $\beta$.
$P(M = \downarrow | \beta)$ can be an asymmetric ($c_s \ne c_{-s}$) or symmetric ($c_s = c_{-s}$) function of $\beta$ depending on the choice of phase angles, unlike in QSPI \cite{SinananSingh2024singleshotquantum} where $P(M = \downarrow | \beta)$ is always a symmetric function, thus limiting its usefulness for asymmetrically distributed parameter detection.

In the absence of any displacement signal, $\beta = 0$, thus no displacement is applied to the oscillator, and the GQSP and its inverse sequence cancel out, implying no change in the initial qubit state. This implies that $\sum_{s = -d}^{d} c_s = 1$ for $\beta = 0$. For cases where $0\notin [\beta_{-th}, \beta_{+th}]$, $P(M = \downarrow| \beta) = 1$ at $\beta=0$ can be treated as a point of exception.

\vspace*{-0.3\baselineskip}
\subsection{Error Scaling and Circuit Depth}\label{sec:results-error}

\vspace*{-1.3\baselineskip}
\begin{figure}[ht]
    \centering
    \includegraphics[width=1\linewidth]{diagrams/GQSPI_Response_ErrorDiag.pdf}
     \vspace*{-1.5\baselineskip}
    \caption{Error analysis of the GQSPI protocol.}
    \label{fig:err-analysis}
    \vspace*{-0.5\baselineskip}
\end{figure}

Now that the qubit measurement probability response function has been expressed as polynomial in Eq. \eqref{eq:csbeta}, the error density of the protocol can be viewed as the error in approximating the ideal function,
\begin{align}
    P_{ideal} = \label{eq:p_ideal}
    \begin{cases}
    1 , \quad \beta_{-th} \leq \beta \leq \beta_{+th}
    \\
    0, \quad \text{otherwise}
\end{cases}.
\end{align}
The probability of decision error density $p_{err}(\beta_{-th}, \beta_{+th}, \kappa)$ of the protocol can be expressed as,
\begin{align}
    &p_{err}(\beta_{-th}, \beta_{+th}, \kappa) 
    = \nonumber
    \frac{\kappa}{\pi} 
    \int_{-\frac{\pi}{2\kappa}}^{\frac{\pi}{2\kappa}} 
    |P_{ideal} - P(M = \downarrow | \beta)|d\beta
    \\
     &= \label{eq:P-err-form}
      \frac{ (\beta_{+th} - \beta_{-th})\kappa}{\pi}
    +
    \sum_{s = -d}^{d} c_s H_s, 
    \\
    &
    \text{where,}~H_s = \text{sinc}(\pi s) + i \frac{e^{i( 2\kappa \beta_{+th}) s} - e^{i( 2\kappa \beta_{-th}) s}}{\pi s}
\end{align}


As shown in Fig. \ref{fig:err-analysis}, assume that the error tolerance of the protocol is $\epsilon$ everywhere expect for $\beta \in [\beta_{-th}-\frac{\sigma}{2}, \beta_{-th}+\frac{\sigma}{2}] \cup [\beta_{+th}-\frac{\sigma}{2}, \beta_{+th}+\frac{\sigma}{2}]$ which are considered as the transition regions of width $\sigma$ around each of the thresholds.
Then the total probability of decision error denoted by $p_{err}$ can be approximately expressed as,
\begin{align}
    p_{err} 
    \approx
    \epsilon \left( \frac{\pi}{\kappa} - 2\sigma \right) + \frac{\sigma}{2}
\end{align}
where we assume that $\epsilon$ and $\sigma$ are of the same order of magnitude. Using the polynomial approximation of a rectangular function result obtained in \cite{low2017hamiltoniansimulationuniformspectral, SinananSingh2024singleshotquantum}, the degree of the polynomial $d$ is related to $p_{err}$ as, 
\begin{align}
    d &\propto \frac{1}{\kappa p_{err}}\log{(\kappa p_{err})}
    \quad
    \text{or equivalently,} \label{eq:degree-perr}
    \quad
    p_{err} \propto \frac{1}{\kappa d}\log{(d)}.
\end{align}


This brings us to concluding the following theorem:
\begin{thm}\label{thm-2}
    An interferometry protocol obtained by applying a degree-$d$ hybrid-GQSP before and after the displacement signal $S_\beta$ acts on an oscillator, realizes the probability of measuring the qubit in ground state $P(M = \downarrow | \beta) = \sum_{s = -d}^{d} c_s e^{i(2\kappa) \beta s}, \kappa\in\mathbb{R}$, achieving arbitrary polynomials w.r.t. $e^{i2\kappa\beta}$, thus improving over QSPI \cite{SinananSingh2024singleshotquantum} which can only realize symmetric (even) functions. Such polynomial transformation produces a desirable interference on the quantum state of the system suitable for binary decision-making with probability of decision error $p_{err}  \propto \mathcal{O}(\frac{1}{\kappa d}\log{d})$. 
\end{thm}

Numerical optimization and analytical methods established in several of the QSP angle finding techniques \cite{chao2020findinganglesquantumsignal, Haah2019product, laneve2025generalizedquantumsignalprocessing} can be used to obtain $\vec{\theta} = \{\theta_0, \theta_1 \cdots \theta_d\}, ~\vec{\phi} = \{\phi_0, \phi_1 \cdots \phi_d\}$, and $\lambda_0$ that create the optimal detection state on the qubit-oscillator system.

The detailed treatment of the GQSPI protocol is presented in Appendix~\ref{sec:appndx-GQSPI} with derivations for $P(M = \downarrow | \beta)$ and $p_{err}(\beta_{-th}, \beta_{+th}, \kappa)$.

\subsection{Displacement Signal Detection for Stochastic Parameter} \label{sec:mp_ta-beta-rv}
Oftentimes the parameter of interest is a random variable. Let $\beta$ follow a probability density function $f_{\beta}(\mu_{\beta}, \sigma^2_{\beta})$ with  mean $\mu_{\beta}$, and variance $\sigma^2_{\beta}$. In this case, the decision error density transforms into an expectation value
\begin{align}
    &\nonumber
    p_{err}(\beta_{-th}, \beta_{+th}, \kappa, \sigma_{\beta}, \mu_{\beta})
     = 
     \frac{\kappa}{\pi}
     \mathbb{E} \left[
    |P_{ideal} - P(M = \downarrow | \beta) |\right ]
    \\
    &\text{where, } \mathbb{E}[ P(M = \downarrow | \beta) ] = \sum_{s = -d}^{d} c_s \left[\int_{-\infty}^{\infty}e^{i(2\kappa)\beta s}
      f_{\beta}(\mu_{\beta}, \sigma^2_{\beta}) d\beta \right]
\end{align}

Further, when actually using the GQSPI protocol, since the only change required is to adjust the decision error density $p_{err}(\beta_{-th}, \beta_{+th}, \kappa)$, therefore, the optimization algorithm minimizing the decision error density can accommodate the prior information of $\beta$ to realize an appropriate polynomial transformation.
The particular case of $\beta$ following the Gaussian probability density is presented in Appendix~\ref{sec:appndx-gaussian-beta}. The current treatment can also be recast for deterministic $\beta$ buried in classical random noise with a known distribution.

\vspace*{-0.3\baselineskip}
\subsection{GQSPI with Dephasing Noise on the Oscillator} 
\label{sec:mp_ta-dephasing}
Similar to qubit decoherence errors, the oscillator also suffers from dephasing, such that a dephasing error occurs during the application of each conditional displacement gate, given by the dephasing operator $R_{osc}(\gamma) = e^{-i\gamma \hat{n}}$. The parameter $\gamma$ could be a random variable drawn from a known distribution, but is usually a fixed value as a coherent error during the duration of the protocol (for shorter circuits).
 This will change $\mathcal{D}_c(i \kappa / \sqrt{2})$ into a noisy version:
    \begin{align}
        \mathcal{D}_c(\alpha) &= \nonumber
        R_{osc}^\dagger(\gamma) \mathcal{D}_c(i \kappa / \sqrt{2}) R_{osc}(\gamma) 
        \\
        &=
        \begin{bmatrix}
        e^{i \kappa (\hat{x} \cos\gamma  + \hat{p} \sin\gamma) } & 0 \\
        0 & e^{-i \kappa (\hat{x} \cos\gamma  + \hat{p} \sin\gamma) }
    \end{bmatrix}
    \end{align}
Then the GQSPI sequence with dephasing error occurring for each iteration is of the form,
    \begin{align}
    G_{\vec{\theta}, \vec{\phi}, \lambda_0}(\omega) 
    &= \nonumber
    \left(\prod_{i=1}^{d} (R(\theta_i, \phi_i, 0)\otimes I_{osc}) \mathcal{D}_c(i \kappa / \sqrt{2}) R_{osc}(\gamma_i) \right)
    \\
    & \nonumber
    \times (R(\theta_0, \phi_0, \lambda_0)\otimes I_{osc})
    \\
    &= \nonumber
    R_{osc}(\sum_{i=1}^d \gamma_i)  \left(\prod_{i=1}^{d} (R(\theta_i, \phi_i, 0)\otimes I_{osc}) \mathcal{D}_c(\alpha'_i) \right)
    \\
    &
    \times
    (R(\theta_0, \phi_0, \lambda_0)\otimes I_{osc}) 
    \end{align}
where,
\begin{align*}
    \alpha'_i = \frac{i\kappa \cos\Gamma_i + \kappa \sin\Gamma_i}{\sqrt{2}} = \frac{\kappa}{\sqrt{2}}e^{i(\frac{\pi}{2} -\Gamma_i)}, 
    \quad
    \Gamma_i = \sum_{j=1}^{i} \gamma_j
\end{align*}
Thus, the dephasing noise rotates each of the oscillator displacements conditioned on the qubit by a varying amount, accumulating all the random rotation angles from the previous iterations.
For such a GQSP set up, the general form of the polynomial 
$P_d(\vec{\gamma})$, where $\vec{\gamma}= \{\gamma_1,\gamma_2 \cdots, \gamma_d\}$, can be expressed as, 
\begin{align}
    P_d(\vec{\gamma}) 
    &= \nonumber
    \sum_{\textbf{s} \in \{-1, 1\}^d} p_{\textbf{s} } \mathcal{D}\left(
    \sum_{k=1}^{d}s_k\alpha'_k 
    \right)
    \\
    & \label{eq:gen-poly-dephasing}
    \times
    \exp{
    \left(
    -i\frac{\kappa^2}{2}
    \sum_{1 \leq l < k \leq d } s_k s_l
    \sin(\Gamma_k - \Gamma_l)
    \right)}, 
\end{align}
where $s_k \in \{-1, 1\}~\forall~k~\in \{1,2, \cdots, d\}$, and $p_{\textbf{s}} \in \mathbb{C}$. $Q_d(\vec{\gamma})$ is also of similar form.
Thus, the polynomial $P$ can be viewed as a linear combination of rotated oscillator displacement operators with varying magnitude and direction.

The overall unitary acting on the qubit-oscillator system taking into account the dephasing noise for the entire interferometry protocol is given by,
 \begin{align}
     &U(\beta, \kappa, \vec{\gamma}) 
     = 
     \nonumber G^{-1}_{d,\vec{\theta}, \vec{\phi}, \lambda_0}(\vec{\gamma}) ~S_{\beta}~ G_{d,\vec{\theta}, \vec{\phi}, \lambda_0}(\vec{\gamma})
     \\
     &= \label{eq:u-deph}
     \begin{bmatrix}
         P^{\dagger}_d(\vec{\gamma})& Q^{\dagger}_d(\vec{\gamma})
         \\
         * & *
     \end{bmatrix}
     I \otimes \mathcal{D}\left( -\frac{\beta}{\sqrt{2}}e^{i\Gamma_d} \right)
     \begin{bmatrix}
         P_d(\vec{\gamma})& *
         \\
         Q_d(\vec{\gamma})& *
     \end{bmatrix}
 \end{align}

The $P(M = \downarrow| \beta)$ obtained for the dephasing noise scenario is of the form,
\begin{align*}
    P(M = \downarrow| \beta) 
   &=
   \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} })
    \\
    &
    \times
    e^{i\Theta_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}}
    e^{-|\alpha_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}|^2/2}
    e^{i(\kappa\beta)\Lambda_{\textbf{s}, \textbf{r}}} 
\end{align*}
where, $e^{i\Theta_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}}$ and $ e^{-|\alpha_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}|^2/2}$ are overall phase and amplitude factors, respectively, obtained from composing the scrambled conditional displacement gates, and $e^{i\kappa\beta \Lambda_{\textbf{s}, \textbf{r}}} $ is comparable to the factor $e^{i 2\kappa \beta s}$ obtained previously in Eq.~\eqref{eq:csbeta}, and \eqref{eq:qspi_eq_20}. Detailed expressions of the parameters can be found in Appendix~\ref{sec:appnx-gqspi-deph}.

It can be shown that for small dephasing rates $\gamma$ (same $\gamma$ in every iteration), $P(M = \downarrow| \beta)$ has an overall rescaling factor of $(1 - \Omega_1\gamma^2 - \Omega_2\beta \gamma^3 + \mathcal{O}(\gamma^4))$ for $\Omega_1, \Omega_2\in\mathbb{R}$, which implies that $P(M = \downarrow| \beta)$ remains close to 1 up to a 2nd-order in error strength $\gamma$ (linear $\gamma$ term vanishes), proving the protocol to be robust against oscillator dephasing for $\beta$ values close to $0$ and around the thresholds with small deviations outside the threshold regions.

\subsection{Simulation Results} \label{sec:results}
 \vspace*{-0.5\baselineskip}
\begin{figure}[ht]
    \centering
    \includegraphics[width=0.9\linewidth]{diagrams/Plots/gqspi136913.png}
     \vspace*{-0.5\baselineskip}
    \caption{Qubit repsonse function for asymmetric thresholding problem. $P(M = \downarrow | \beta)$ plotted against $\beta$ for $\kappa = \frac{1}{2048}$ for degrees $1,3, 6, 9, \text{and }13$ with the blue dotted lines representing the thresholds at $\beta_{-th} = -\frac{\pi}{8\kappa}$ and $\beta_{+th} = \frac{\pi}{4\kappa}$. }
    \label{fig:GQSPI-Sim-Results}
    \vspace*{-0.5\baselineskip}
\end{figure}

From the simulation results shown in Fig. \ref{fig:GQSPI-Sim-Results}, it can be observed that $P(M = \downarrow| \beta)$ is asymmetric with respect to $\beta = 0$ for GQSPI degrees greater than $1$ as expected. For degree $d=1$, the protocol reduces to a symmetric polynomial as shown in Appendix \ref{sec:gqspi-degree1-symmetry}. The transition becomes sharper with increasing polynomial degree, and the total decision error decreases with increasing GQSP degree, as seen in Fig. \ref{fig:perr_vs_deg}, roughly following the expected behavior derived in Eq. \eqref{eq:degree-perr}.

\begin{figure}[ht]
    \centering
\includegraphics[width=0.8\linewidth]{diagrams/Plots/perr_vs_deg.png}
    \vspace*{-0.5\baselineskip}
    \caption{Total probability of decision error $p_{err}$ vs GSQPI degree obtained from loss calculated for simulation results in Fig. \ref{fig:GQSPI-Sim-Results}. The red dots on log-log plot indicate the $p_{err}$ value for the given GQSP degree. The trendlines for $\frac{1}{d}, \frac{1}{d^\alpha}$ and fitting line for $\frac{1}{d}\log{d}$ are plotted against the given GQSP degrees to show the general trend of non-linear error suppression with increasing degree.}
    \label{fig:perr_vs_deg}
    \vspace*{-0.5\baselineskip}
\end{figure}


In certain cases where a displacement signal must be detected over multiple bands of values, several thresholds come into the picture. Assuming a simple case of two non-overlapping ranges of thresholds, say $[\beta_{1},\beta_{2}]$, and $[\beta_{3},\beta_{4}]$, simulation results obtained in Fig. \ref{fig:double-window} show that GQSPI can be used for multi-thresholding problems as well. The expression for $p_{err}(\beta_{1}, \beta_{2}, \beta_{3}, \beta_{4}, \kappa)$ for the two bands of thresholds can be found in Appendix~\ref{sec:multi-threshold} Eq.~\eqref{eq:multi-thesh-p_err} and the general case of $m$ non-overlapping thresholds is shown in Eq.~\eqref{eq:general-multi-thresh-perr}. 
It should be noted that such cases require higher GQSP degrees due to the presence of multiple narrow transition regions.

Code for the simulation results obtained in this section can be found in \cite{code_repo_gqspi}.
\begin{figure}[ht]
    \centering
    \includegraphics[width=0.9\linewidth]{diagrams/Plots/DoubleThresh.png}
    \vspace*{-0.8\baselineskip}
    \caption{Qubit response function for the multi-thresholding problem. $P(M = \downarrow | \beta)$ plotted against $\beta$ for $\kappa = \frac{1}{2048}$ for degrees $9, 13, \text{ and }15$ with the blue dotted lines representing the thresholds at $\beta_{1} = -0.75\frac{\pi}{2\kappa}$, $\beta_{2} = -0.45\frac{\pi}{2\kappa}, \beta_{3} = -0.125\frac{\pi}{2\kappa}$, $\beta_{4} = 0.75\frac{\pi}{2\kappa}$.
    The region for which we wish to detect $\beta$ is $[\beta_{1},\beta_{2}] \cup [\beta_{3},\beta_{4}]$.
    }
    \label{fig:double-window}
    \vspace*{-0.5\baselineskip}
\end{figure}

\section{Conclusions} \label{sec:disc}
The ability to actively detect a general quantum displacement signal over a single or multiple threshold bands in the presence of classical and quantum noise is a challenging practical problem. We address these aspects of quantum detection problem using the proposed GQSPI framework that is capable of systematically improving the decision accuracy while being robust against quantum dephasing noise. By viewing the displacement signal detection problem from a polynomial approximation perspective, we uncover deeper connections between quantum detection theory and quantum algorithms, thereby opening an exciting avenue for binary decision problems in the quantum domain.

\section{Acknowledgment}
This work is supported by the U.S. Department of Energy, Office of Science, Advanced Scientific Computing Research, under contract number DE-SC0025384.

% \newpage
% \balance
\bibliographystyle{unsrt}
\bibliography{ref}

% \newpage
\input{appendix}

\end{document}
\onecolumn
\appendices

\renewcommand{\thesubsection}{\thesection-\Roman{subsection}}
\makeatletter
\def\@seccntformat#1{%
  \ifcsname the#1\endcsname
    \csname the#1\endcsname\hskip 0.75em
  \fi}
\makeatother

\section{Generalized Quantum Signal Processing for Hybrid Qubit-Oscillator Systems} \label{sec:appndx-GQSP-Poly-Forms}

\subsection{Nature of Polynomials}

Assume  $\hat{\omega} = e^{i\kappa \hat{x}}$ for simplicity of calculation.
\\
For GQSP of degree 1, the polynomials obtained by simply multiplying one iteration of the qubit rotation gates and the conditional displacement gate results in the following polynomials,
\begin{align}
    P_1(\hat{x}) &= \label{eq:P1}
    e^{i(\phi_{1}+ \lambda_0)} \left(
        e^{i \phi_0} \cos{\theta_{1}} \cos{\theta_0} e^{i \kappa \hat{x}}
        +
       \sin{\theta_{1}}\sin{\theta_0} e^{-i \kappa \hat{x}}
        \right) 
        \\
        Q_1(\hat{x}) &= \label{eq:Q1}
    e^{i\lambda_0} \left(
        e^{i\phi_0} \sin{\theta_{1}}   \cos{\theta_0} e^{i \kappa \hat{x}}
        -
        \cos{\theta_{1}}\sin{\theta_0} e^{-i \kappa \hat{x}}
        \right)
\end{align}

It can be observed that the coefficients of the polynomial can be complex or real, depending on the choice of $\vec{\theta} = \{\theta_0, \theta_1 \cdots \theta_d\}, ~\vec{\phi} = \{\phi_0, \phi_1 \cdots \phi_d\}$, and $\lambda_0$.


Going from degree $d$ to $d+1$ involves multiplying with the following ``\textbf{unit}":
\begin{align}
    (R(\theta_i, \phi_i, 0)\otimes I_{osc})D_c(i\kappa/\sqrt{2}) 
    &= 
    \label{eq:fwd-unit}
    \begin{bmatrix} 
        e^{i \phi_{d+1}}\cos{\theta_{d+1}}e^{i \kappa \hat{x}} & e^{i\phi_{d+1}}\sin{\theta_{d+1}}e^{-i \kappa \hat{x}} \\
        \sin{\theta_{d+1}}e^{i \kappa \hat{x}} & 
        -\cos{\theta_{d+1}}e^{-i \kappa \hat{x}}
    \end{bmatrix}
\end{align}

Thus the recursive relationship between the polynomials are: 
\begin{align}
    P_{d+1}(\hat{x}) 
    &= \label{eq:Pd-recurr}
        e^{i \phi_{d+1}}\cos{\theta_{d+1}}e^{i \kappa \hat{x}}P_d(\hat{x}) 
        + 
        e^{i\phi_{d+1}}\sin{\theta_{d+1}}e^{-i \kappa \hat{x}} Q_d(\hat{x})
        \\
    Q_{d+1}(\hat{x}) 
    &= \label{eq:Qd-recurr}
        \sin{\theta_{d+1}}e^{i \kappa \hat{x}}P_d(\hat{x}) -
        \cos{\theta_{d+1}}e^{-i \kappa \hat{x}}Q_d(\hat{x}) 
\end{align}

Combining the Eq. \eqref{eq:Pd-recurr}, \eqref{eq:Qd-recurr} and Eq. \eqref{eq:P1}, \eqref{eq:Q1}, and calculating the polynomial $P$ for degree 2 and 3,
\begin{align*}
    P_{1}(\hat{x}) &= p_1 e^{i \kappa \hat{x}} + p_{-1} e^{-i \kappa \hat{x}} 
    \\
    P_{2}(\hat{x}) &= p_{-2}e^{-i 2 \kappa \hat{x}} + p_{2}e^{i 2 \kappa \hat{x}} + p_0
    \\
    P_{3}(\hat{x}) &= p_{3} e^{i 3\kappa \hat{x}} + p_{-3} e^{-i 3\kappa \hat{x}} + p_{-1} e^{-i \kappa \hat{x}} + p_{1} e^{i \kappa \hat{x}}
\end{align*}

Thus, the general form of the polynomial is 
\begin{align}
    P_d(\hat{x}) = \sum_{n=-d}^{d} p_n \hat{x}^n, \quad n = \{-d, -d+2, \cdots, d-2, d\}.
    \label{eq:P-general}
\end{align}
This polynomial form is obtained for any operator $\hat\omega$.

\subsection{Recursive Relationship Between the GQSP Coefficients}

Expanding both sides of Eq. \eqref{eq:Pd-recurr} using the general form of polynomial given in Eq \eqref{eq:P-general}, 
\begin{align*}
    \sum_{r=-(d+1)}^{d+1} p^{(d+1)}_r e^{i r \kappa \hat{x}}
    &= 
    e^{i \phi_{d+1}}\cos{\theta_{d+1}}e^{i \kappa \hat{x}}
    \left(\sum_{r=-(d)}^{d} p^{(d)}_re^{i r \kappa \hat{x}}
    \right)
    + e^{i\phi_{d+1}}\sin{\theta_{d+1}}e^{-i \kappa \hat{x}} 
        \left(\sum_{r=-(d)}^{d} q^{(d)}_re^{i r \kappa \hat{x}}
        \right)
    \\
    &=
    e^{i \phi_{d+1}}\cos{\theta_{d+1}}
    \left( p^{(d)}_de^{i (d+1) \kappa \hat{x}} + p^{(d)}_{d-2}e^{i (d-1) \kappa \hat{x}} +
    \cdots
    +
     p^{(d)}_{-d}e^{i (-d+1) \kappa \hat{x}}
    \right)
    \\
    &+
    e^{i\phi_{d+1}}\sin{\theta_{d+1}}
    \left(
     q^{(d)}_de^{i (d-1) \kappa \hat{x}} + q^{(d)}_{d-2}e^{i (d-3) \kappa \hat{x}} +
    \cdots
    +
     q^{(d)}_{-d}e^{-i (d+1) \kappa \hat{x}}
    \right)
\end{align*}
Comparing coefficients on both sides results in the following relations,
\begin{align*}
    p^{(d+1)}_r = 
    \begin{cases}
        e^{i\phi_{d+1}} \cos{\theta_{d+1}}p^{(d)}_{r-1}, 
        \quad r = d+1
        \\
        e^{i\phi_{d+1}} \sin{\theta_{d+1}}q^{(d)}_{r+1},
        \quad r = -(d+1)
        \\
        e^{i\phi_{d+1}}\cos{\theta_{d+1}}p^{(d)}_{r-1}
        + 
        e^{i\phi_{d+1}}\sin{\theta_{d+1}}q^{(d)}_{r+1}, \quad |r| \le (d-1)
    \end{cases},
\end{align*}
Using a similar approach for coefficients of the polynomial $Q$, 
\begin{align*}
    q^{(d+1)}_r = 
    \begin{cases}
        \sin{\theta_{d+1}}p^{(d)}_{r-1}, 
        \quad r = d+1
        \\
        -\cos{\theta_{d+1}}q^{(d)}_{r+1},
        \quad r = -(d+1)
        \\
        \sin{\theta_{d+1}}p^{(d)}_{r-1}
        -
        \cos{\theta_{d+1}}q^{(d)}_{r+1}, \quad |r| \le (d-1)
    \end{cases}.
\end{align*}


\section{Generalized Quantum Signal Processing Interferometry Protocol} \label{sec:appndx-GQSPI}

\subsection{Structure of \texorpdfstring{$U(\beta, \kappa)$}{}}

$U(\beta, \kappa)$ is the overall unitary obtained in the protocol by sandwiching the displacement signal $S_{\beta}$ by a $d$-degree GQSPI and its inverse,
\begin{align}
     U(\beta, \kappa) \nonumber
     &= \nonumber G^{-1}_{d,\vec{\theta}, \vec{\phi}, \lambda_0}(\omega) ~S_{\beta}~ G_{d,\vec{\theta}, \vec{\phi}, \lambda_0}(\omega)\\
     &= \nonumber
     \begin{bmatrix}
        \sum_{n=-(d)}^{d} p_n^{*} e^{-i n\kappa \hat{x}} 
        &
         \sum_{n=-(d)}^{d} q_n^{*} e^{-i n\kappa \hat{x}}
         \\
        -\sum_{n=-(d)}^{d} q_n e^{i n\kappa \hat{x}}
        &
         \sum_{n=-(d)}^{d} p_n e^{i n\kappa \hat{x}}
    \end{bmatrix}
    \begin{bmatrix}
        e^{i \beta \hat{p}}  & 0 \\
        0 & e^{i \beta \hat{p}}
    \end{bmatrix}
    \begin{bmatrix}
        \sum_{n=-(d)}^{d} p_n e^{i n\kappa \hat{x}} 
        &
         -\sum_{n=-(d)}^{d} q^{*}_n e^{-i n\kappa \hat{x}}
         \\
        \sum_{n=-(d)}^{d} q_n e^{i n\kappa \hat{x}}
        &
         \sum_{n=-(d)}^{d} p^{*}_n e^{-i n\kappa \hat{x}}
    \end{bmatrix}
     \\
     &= \label{eq:U-beta}
     \sum_{n,m=-d}^{d} C_{nm} e^{- i n \kappa \hat{x}} S_{\beta} e^{i m \kappa \hat{x}}, \quad \text{where }
     C_{nm} = 
    \begin{bmatrix}
        p^{*}_n p_m +  q^{*}_n q_m
        &
        -  \left(
        p^{*}_n q^{*}_m 
        - q^{*}_n p^{*}_m
        \right)
        \\
        \left( 
        p_n q_m
        -
        q_n p_m 
        \right)
        &
        q_n q^{*}_m +  p_n p^{*}_m
    \end{bmatrix}
\end{align}
where $n,m = \{-d, -d+2, \cdots, d-2, d\}$.

The top-left element of the operator $U(\beta, \kappa)$ can be expressed as,

\begin{align}
    U(\beta, \kappa)_{00} &= \label{eq:u00}
    \sum_{n,m=-d}^{d} \left( p^{*}_n p_m +  q^{*}_n q_m \right)e^{-i n\kappa \hat{x}} e^{i\beta \hat{p}} e^{i m \kappa \hat{x}}
    \\
    &= \nonumber
    \sum_{n,m=-d}^{d} \left( p^{*}_n p_m +  q^{*}_n q_m \right)
    e^{i\beta \hat{p}} e^{-i n\kappa \hat{x}} e^{-i n\kappa \beta}  e^{i m \kappa \hat{x}}
    % \\
    % &= \nonumber
    % \sum_{n,m=-d}^{d} \left( p^{*}_n p_m +  q^{*}_n q_m \right)
    % e^{i\beta \hat{p}}e^{-i n\kappa (\hat{x} -\beta)} e^{i m \kappa \hat{x}}
    % \\
    % &= \nonumber
    % e^{i\beta \hat{p}} 
    % \left[ 
    %  \sum_{n,m=-d}^{d} 
    %   p^{*}_n e^{-i n\kappa (\hat{x} -\beta)} p_m e^{i m \kappa \hat{x}} 
    %   +  q^{*}_n e^{-i n\kappa (\hat{x} -\beta)} 
    %   q_m e^{i m \kappa \hat{x}}
    % \right]
    % \\
    % &= e^{i\beta \hat{p}} 
    %  \left[ 
    %  P^{*}_{d}(\hat{x}-\beta)P_{d}(\hat{x}) +
    %  Q^{*}_{d}(\hat{x}-\beta)Q_{d}(\hat{x})
    % \right]
\end{align}

% \subsection{Form of \texorpdfstring{$U^{\dagger}(\beta, \kappa)_{00}U(\beta, \kappa)_{00}$}{}} \label{sec:u-beta00u-inv00}

% \begin{align}
%     U^{\dagger}(\beta, \kappa)_{00}U(\beta, \kappa)_{00}
%     &= \nonumber
%     \left(\sum_{n,m=-d}^{d} \left( p^{*}_n p_m +  q^{*}_n q_m \right)e^{-i n\kappa \hat{x}} e^{i\beta \hat{p}} e^{i m \kappa \hat{x}}\right)^{\dagger}
%     \left(
%     \sum_{n',m'=-d}^{d} \left( p^{*}_{n'} p_{m'} +  q^{*}_{n'} q_{m'} \right)e^{-i n' \kappa \hat{x}} e^{i\beta \hat{p}} e^{i m' \kappa \hat{x}}
%     \right)
%     \\
%     &= \nonumber
%     \left( \sum_{n,m=-d}^{d} \left( p_n p^{*}_m +  q_n q^{*}_m \right)e^{-i m\kappa \hat{x}} e^{-i\beta \hat{p}} e^{i n \kappa \hat{x}}
%     \right)
%     \left(
%     \sum_{n',m'=-d}^{d} \left( p^{*}_{n'} p_{m'} +  q^{*}_{n'} q_{m'} \right)e^{-i n' \kappa \hat{x}} e^{i\beta \hat{p}} e^{i m' \kappa \hat{x}}
%     \right)
%     \\
%     &= \nonumber
%     \sum_{n, m, n', m' = -d}^{d} 
%     \left( p_n p^{*}_m +  q_n q^{*}_m \right)\left( p^{*}_{n'} p_{m'} +  q^{*}_{n'} q_{m'} \right)
%     e^{-i m\kappa \hat{x}} e^{-i\beta \hat{p}} e^{ i n \kappa \hat{x}} e^{-i n' \kappa \hat{x}} e^{i\beta \hat{p}} e^{i m' \kappa \hat{x}}
%     \\
%     &= \nonumber
%     \sum_{n, m, n', m' = -d}^{d} 
%     \left( p_n p^{*}_m +  q_n q^{*}_m \right)\left( p^{*}_{n'} p_{m'} +  q^{*}_{n'} q_{m'} \right)
%     e^{-i m\kappa \hat{x}} e^{-i\beta \hat{p}} e^{i (n - n') \kappa \hat{x}} e^{i\beta \hat{p}} e^{i m' \kappa \hat{x}}
%     \\
%     &= \nonumber
%     \sum_{n, m, n', m' = -d}^{d} 
%     \left( p_n p^{*}_m +  q_n q^{*}_m \right)\left( p^{*}_{n'} p_{m'} +  q^{*}_{n'} q_{m'} \right)
%     e^{-i m\kappa \hat{x}} e^{-i\beta \hat{p}}  e^{i\beta \hat{p}} e^{i (n - n') \kappa \hat{x}} e^{-i(n - n')\kappa \beta} e^{i m' \kappa \hat{x}} 
%     \\
%     &= \nonumber
%     \sum_{n, m, n', m' = -d}^{d} 
%     \left( p_n p^{*}_m +  q_n q^{*}_m \right)\left( p^{*}_{n'} p_{m'} +  q^{*}_{n'} q_{m'} \right) e^{-i(n - n')\kappa \beta}
%     e^{i (- m + n - n' + m') \kappa \hat{x}} 
%     \\
%     &= \label{eq:u_dag00u00}
%     \sum_{n, m, n', m' = -d}^{d} 
%     \left( p_n p^{*}_m +  q_n q^{*}_m \right)\left( p^{*}_{n'} p_{m'} +  q^{*}_{n'} q_{m'} \right) e^{-i(n - n')\kappa \beta}
%     e^{i (n - n' - m + m') \kappa \hat{x}} 
% \end{align}



\subsection{Qubit Probability Response Function \texorpdfstring{$P(M = \downarrow| \beta)$}{}} 
\label{sec:prob-beta}

Using the equation obtained in Eq.~\eqref{eq:u00}, the qubit probability response function $P(M = \downarrow| \beta)$ can be expressed as,
\begin{align}
    P(M = \downarrow| \beta) 
    &= \nonumber
   \langle 0|_{osc} U^{\dagger}(\beta, \kappa)_{00}U(\beta, \kappa)_{00}|0 \rangle _{osc}
   \\
   &= \nonumber
   \int_{-\infty}^{\infty} 
   \left[
   \sum_{n, m, n', m' = -d}^{d} 
    \left( p_n p^{*}_m +  q_n q^{*}_m \right)\left( p^{*}_{n'} p_{m'} +  q^{*}_{n'} q_{m'} \right) e^{-i(n - n')\kappa \beta}
    e^{i (n - n' - m + m') \kappa \hat{x}}  \right]
   \psi^{2}_0(x) dx
   \\
  &=  \nonumber
  \sum_{n, m, n', m' = -d}^{d} 
  \left[
    \left( p_n p^{*}_m +  q_n q^{*}_m \right)\left( p^{*}_{n'} p_{m'} +  q^{*}_{n'} q_{m'} \right) 
      e^{-i(n - n')\kappa \beta}
      \int_{-\infty}^{\infty}
    e^{i (n - n' - m + m') \kappa x} 
   \psi^{2}_0(x) dx 
   \right] \\
   &= \nonumber
   \sum_{n, m, n', m' = -d}^{d} 
  \left[
    \left( p_n p^{*}_m +  q_n q^{*}_m \right)\left( p^{*}_{n'} p_{m'} +  q^{*}_{n'} q_{m'} \right) 
     e^{-\frac{1}{4} k^2 (n - n' - m + m')^2} \right] e^{-i(n - n')\kappa \beta} 
    %  \\
    %  &= \nonumber 
    %  \sum_{n, m, n', m' = -d}^{d} 
    % \left[
    % \left( p_n p^{*}_m +  q_n q^{*}_m \right)\left( p^{*}_{n'} p_{m'} +  q^{*}_{n'} q_{m'} \right) 
    %   e^{-\frac{1}{4} k^2 (n - n' - m + m')^2} \right] e^{i(n' - n)\kappa \beta}
     \\
     &= \nonumber
      \sum_{s = -d}^{d} 
      \left(
      \sum_{n, m, r = -d}^{d}
     \left[
    \left( p_n p^{*}_{m} +  q_n q^{*}_{m} \right)\left( p^{*}_{n + 2s} p_{m + 2r} +  q^{*}_{n + 2s} q_{m + 2r} \right) 
     e^{- k^2 (r - s )^2} \right] 
      \right)e^{i(2\kappa \beta) s}
      \\
      &= 
      \sum_{s = -d}^{d} c_s \nu^s(\beta), 
      \quad \text{where }
       c_s =
      \sum_{n, m, r = -d}^{d}
    \left( p_n p^{*}_{m} +  q_n q^{*}_{m} \right)\left( p^{*}_{n + 2s} p_{m + 2r} +  q^{*}_{n + 2s} q_{m + 2r} \right) 
     e^{- k^2 (r - s)^2}
\end{align}

Here we have performed the variable substitution $n' - n = 2s$ and $m' - m = 2r$  with  $s, r \in [-d, d]$. Thus, $n' \rightarrow n + 2s$, and $m' \rightarrow 2r + m$.


We use the result $\int_{-\infty}^{\infty} e^{bx-ax^2} dx= e^{-b^2/4a}\sqrt{\frac{\pi}{a}} 
\rightarrow 
\int_{-\infty}^{\infty} e^{-i(m + n' - m' - n) \kappa x  -x^2}/\sqrt{\pi}dx   = e^{\frac{[-i(m + n' - m' - n) \kappa]^2}{4}} $.

$\psi_0(x) = \pi^{-1/4}e^{-x^2/2}$ is the vacuum state of the oscillator $\Rightarrow \psi^2_0(x) = e^{-x^2}/\sqrt{\pi}$.


\subsection{\texorpdfstring{$P_{ideal}$}{} and Decision Error Density \texorpdfstring{$p_{err}(\beta_{-th}, \beta_{+th}, \kappa)$}{} for Asymmetric \texorpdfstring{$\beta$}{}} 
\label{sec:prob-err-density}

The ideal qubit probability response function as given in Eq.~\eqref{eq:p_ideal},
\begin{align}
    P_{ideal} = \begin{cases}
    1 , \quad \beta_{-th} \leq \beta \leq \beta_{+th}
    \\
    0, \quad \text{otherwise}
\end{cases}
\end{align}

The decision error density $p_{err}(\beta_{-th}, \beta_{+th}, \kappa)$ would be the average of the difference between $P_{ideal}$ and the polynomial approximation obtained from the GQSPI protocol $P(M = \downarrow | \beta)$. The Eq. \eqref{eq:P-err-form} can be obtained as follows:

\begin{align}
    &p_{err}(\beta_{-th}, \beta_{+th}, \kappa) 
    = \nonumber
    \frac{\kappa}{\pi} 
    \int_{-\frac{\pi}{2\kappa}}^{\frac{\pi}{2\kappa}} 
    |P_{ideal} - P(M = \downarrow | \beta)|d\beta
    \\
    &= 
    \frac{\kappa}{\pi} 
    \left[
    \int_{-\frac{\pi}{2\kappa}}^{\beta_{-th}} P(M = \downarrow | \beta) d\beta
    +
     \int_{\beta_{-th}}^{\beta_{+th}} |1 - P(M = \downarrow | \beta)|d\beta
     +
      \int_{\beta_{+th}}^{\frac{\pi}{2\kappa}} P(M = \downarrow | \beta) d\beta
    \right]
    \\
    &= \nonumber
    \frac{\kappa}{\pi} 
    \left[
    \int_{-\frac{\pi}{2\kappa}}^{\beta_{-th}} 
    \left(
     \sum_{s = -d}^{d} c_s e^{i(2\kappa \beta) s}
    \right) d\beta
    +
     \int_{\beta_{-th}}^{\beta_{+th}} \left[
     1 - \left(
     \sum_{s = -d}^{d} c_s e^{i(2\kappa \beta) s}
    \right)
    \right]
    d\beta
     +
      \int_{\beta_{+th}}^{\frac{\pi}{2\kappa}}\left(
     \sum_{s = -d}^{d} c_s e^{i(2\kappa \beta) s}
    \right) d\beta
    \right]
    % \\
    % &= \nonumber
    % \frac{\kappa}{\pi} 
    % \left[
    %  \sum_{s = -d}^{d} c_s
    %  \int_{-\frac{\pi}{2\kappa}}^{\beta_{-th}} 
    %  e^{i(2\kappa \beta) s} 
    %  d\beta
    % +
    % \int_{\beta_{-th}}^{\beta_{+th}} 
    % \left[
    % 1 - \left(
    %  \sum_{s = -d}^{d} c_s e^{i(2\kappa \beta) s}
    % \right)
    % \right]
    % d\beta
    % +
    %   \sum_{s = -d}^{d} c_s
    %   \int_{\beta_{+th}}^{\frac{\pi}{2\kappa}}
    %  e^{i(2\kappa \beta) s}
    %  d\beta
    %  \right]
    %  \\
    %  &=\nonumber
    %  \frac{\kappa}{\pi} 
    % [
    % \sum_{s = -d}^{d} c_s 
    %  (\frac{-i}{2 \kappa s})(e^{i( 2\kappa \beta_{-th}) s} - e^{-i \pi s})    +
    %  (\beta_{+th} - \beta_{-th}) - \sum_{s = -d}^{d} c_s 
    %  (\frac{-i}{2 \kappa s})(e^{i( 2\kappa \beta_{+th}) s} - e^{i( 2\kappa \beta_{-th}) s})
    %  \\ \nonumber
    %  &+
    % \sum_{s = -d}^{d} c_s 
    %  (\frac{-i}{2 \kappa s})(e^{i \pi s} - e^{i( 2\kappa \beta_{+th}) s})
    % ]
    % \\
    % &=\nonumber
    % \frac{\kappa}{\pi} 
    % \left[
    % (\beta_{+th} - \beta_{-th})
    % +
    % (\frac{-i}{2\kappa})\sum_{s = -d}^{d} \frac{c_s}{s}
    % \left(
    % 2 [e^{i( 2\kappa \beta_{-th}) s} - e^{i( 2\kappa \beta_{+th}) s}]
    % + (e^{i \pi s} - e^{-i \pi s} )
    % \right)
    %  \right]
    %  \\
    %  &= \nonumber
    % \frac{ (\beta_{+th} - \beta_{-th})\kappa}{\pi}
    % +
    % \sum_{s = -d}^{d}
    % \frac{c_s}{s}\frac{\kappa}{\pi}(\frac{-i}{2\kappa}) 2i \sin{(\pi s)}
    % +
    %  \sum_{s = -d}^{d}
    % \frac{c_s}{s}\frac{\kappa}{\pi}(\frac{-i}{2\kappa}) 2 [e^{i( 2\kappa \beta_{-th}) s} - e^{i( 2\kappa \beta_{+th}) s}]
    \\
    &=
    \nonumber
    \frac{ (\beta_{+th} - \beta_{-th})\kappa}{\pi}
    +
    \sum_{s = -d}^{d} c_s \text{sinc}(\pi s)
    +
     i \sum_{s = -d}^{d} c_s 
     \frac{e^{i( 2\kappa \beta_{+th}) s} - e^{i( 2\kappa \beta_{-th}) s}}{\pi s}
     \\
     &= \nonumber
      \frac{ (\beta_{+th} - \beta_{-th})\kappa}{\pi}
    +
    \sum_{s = -d}^{d} c_s H_s, 
    \quad \quad \text{where}~H_s = \text{sinc}(\pi s) + i \frac{e^{i( 2\kappa \beta_{+th}) s} - e^{i( 2\kappa \beta_{-th}) s}}{\pi s}
\end{align}



\subsection{Multiple Threshold Problem} \label{sec:multi-threshold}

In the example case of detecting $\beta$ over two non-overlapping ranges of thresholds, say $[\beta_{1},\beta_{2}]$, $[\beta_{3},\beta_{4}]$, then, only changes are in the $p_{err}$ calculation. The ideal function in this scenario is,
\begin{align}
    P_{ideal} = \begin{cases}
    1 , \quad \beta_{1} \leq \beta \leq \beta_{2} \cup \beta_{3} \leq \beta \leq \beta_{4}
    \\
    0, \quad \text{otherwise}
\end{cases}
\end{align}
\begin{figure}[h]
    \centering
    \includegraphics[scale=0.6]{diagrams/Multi-Threshdrawio.pdf}
    \caption{Example case of two bands of thresholds}
    \label{fig:gqspi-response-double}
\end{figure}
Taking a similar approach as shown in the previous Sec.~\ref{sec:prob-err-density},
\begin{align}
    &p_{err}(\beta_{1}, \beta_{2}, \beta_{3}, \beta_{4}, \kappa) 
    = \nonumber
    \frac{\kappa}{\pi} 
    \int_{-\frac{\pi}{2\kappa}}^{\frac{\pi}{2\kappa}} 
    |P_{ideal} - P_{approx}(\beta)|d\beta
    \\
    &= \nonumber
    \frac{\kappa}{\pi} 
    \left[
    \int_{-\frac{\pi}{2\kappa}}^{\beta_{1}} P(M = \downarrow| \beta) d\beta
    +
     \int_{\beta_{1}}^{\beta_{2}} |1 - P(M = \downarrow| \beta)|d\beta
     +
      \int_{\beta_{2}}^{\beta_{3}} P(M = \downarrow| \beta) d\beta
      +
     \int_{\beta_{3}}^{\beta_{4}} |1 - P(M = \downarrow| \beta)|d\beta
     +
     \int_{\beta_{4}}^{\frac{\pi}{2\kappa}} P(M = \downarrow| \beta) d\beta
    \right]
    \\
    &= \nonumber
     \frac{\kappa}{\pi} 
    \left[
    \int_{-\frac{\pi}{2\kappa}}^{\beta_{1}} \left(
     \sum_{s = -d}^{d} c_s e^{i(2\kappa \beta) s}
    \right) d\beta
    +
     \int_{\beta_{1}}^{\beta_{2}} |1 - \left(
     \sum_{s = -d}^{d} c_s e^{i(2\kappa \beta) s}
    \right)|d\beta
    +
      \int_{\beta_{2}}^{\beta_{3}} \left(
     \sum_{s = -d}^{d} c_s e^{i(2\kappa \beta) s}
    \right) d\beta
    \right.
    \\ \nonumber
    &\left.
      +
     \int_{\beta_{3}}^{\beta_{4}} |1 - \left(
     \sum_{s = -d}^{d} c_s e^{i(2\kappa \beta) s}
    \right)|d\beta
     +
     \int_{\beta_{4}}^{\frac{\pi}{2\kappa}} \left(
     \sum_{s = -d}^{d} c_s e^{i(2\kappa \beta) s}
    \right) d\beta
    \right]
    % \\
    % &=
    %  \frac{\kappa}{\pi} 
    % \left[
    %  \sum_{s = -d}^{d} c_s 
    %  \left(
    %  \int_{-\frac{\pi}{2\kappa}}^{\beta_{1}} e^{i(2\kappa \beta) s} d\beta
    % \right) 
    % +
    %  \int_{\beta_{1}}^{\beta_{2}} d\beta - 
    %  \sum_{s = -d}^{d} c_s  
    %  \left(
    %  \int_{\beta_{1}}^{\beta_{2}}e^{i(2\kappa \beta) s} d\beta
    % \right)
    % +
    %  \sum_{s = -d}^{d} c_s 
    %  \left(
    %  \int_{\beta_{2}}^{\beta_{3}} e^{i(2\kappa \beta) s}  d\beta
    % \right)
    %  \right.
    % \\
    % &\left.
    %   +
    %  \int_{\beta_{3}}^{\beta_{4}} d\beta - 
    %  \sum_{s = -d}^{d} c_s \left(
    %  \int_{\beta_{3}}^{\beta_{4}}e^{i(2\kappa \beta) s} d\beta
    % \right)
    %  +
    %  \sum_{s = -d}^{d} c_s 
    %   \left(
    %   \int_{\beta_{4}}^{\frac{\pi}{2\kappa}} 
    %  e^{i(2\kappa \beta) s} d\beta
    % \right) 
    % \right]
    % \\
    % &=
    % \frac{\kappa}{\pi} 
    % \left[
    %  \sum_{s = -d}^{d} c_s 
    %  \left(
    %  \int_{-\frac{\pi}{2\kappa}}^{\beta_{1}} e^{i(2\kappa \beta) s} d\beta
    %  -
    %  \int_{\beta_{1}}^{\beta_{2}}e^{i(2\kappa \beta) s} d\beta
    %  +
    %  \int_{\beta_{2}}^{\beta_{3}} e^{i(2\kappa \beta) s}  d\beta
    %  -
    %   \int_{\beta_{3}}^{\beta_{4}}e^{i(2\kappa \beta) s} d\beta
    % +
    % \int_{\beta_{4}}^{\frac{\pi}{2\kappa}} 
    %  e^{i(2\kappa \beta) s} d\beta
    % \right) 
    % \right.
    % \\
    % &\left.
    % +
    %  \int_{\beta_{1}}^{\beta_{2}} d\beta 
    % +
    %  \int_{\beta_{3}}^{\beta_{4}} d\beta
    %  \right]
     % \\
     % &=
     % \frac{\kappa}{\pi} (\beta_{4} - \beta_{3} + \beta_{2} - \beta_{1})
     % +
     % \frac{\kappa}{\pi}
     % \sum_{s = -d}^{d} c_s (\frac{-i}{2 \kappa s})
     % \left[
     % (e^{i( 2\kappa \beta_{1}) s} - e^{-i \pi s})
     % -
     % (e^{i( 2\kappa \beta_{2}) s} - e^{i( 2\kappa \beta_{1}) s})
     % +
     % (e^{i( 2\kappa \beta_{3}) s} - e^{i( 2\kappa \beta_{2}) s})
     % \right.
     % \\
     % &=
     %  \frac{\kappa}{\pi} (\beta_{4} - \beta_{3} + \beta_{2} - \beta_{1})
     %  +
     % \frac{\kappa}{\pi}
     % \sum_{s = -d}^{d} c_s (\frac{-i}{2 \kappa s})
     % \left[
     % 2i \sin(\pi s)
     % +
     % 2(e^{i( 2\kappa \beta_{1}) s} - 
     % e^{i( 2\kappa \beta_{2}) s} )
     % +
     % 2(e^{i( 2\kappa \beta_{3}) s} - 
     % e^{i( 2\kappa \beta_{4}) s} )
     % \right]
     \\
     &= \label{eq:multi-thesh-p_err}
     \frac{\kappa}{\pi} (\beta_{4} - \beta_{3} + \beta_{2} - \beta_{1})
     +
     \sum_{s = -d}^{d} c_s 
     \left[
     \text{sinc}(\pi s)
     +
     (\frac{i}{\pi s}) (e^{i( 2\kappa \beta_{2}) s} - 
     e^{i( 2\kappa \beta_{1}) s} 
     +
     e^{i( 2\kappa \beta_{4}) s} - 
     e^{i( 2\kappa \beta_{3}) s} )
     \right]
\end{align}

This form of $p_{err}(\beta_{1}, \beta_{2}, \beta_{3}, \beta_{4}, \kappa)$ obtained in Eq.~\eqref{eq:multi-thesh-p_err} can also be generalized for more than 2 bands of thresholds. 
Assume that $\beta$ is required to be detected for $m$ number of non-overlapping bands of thresholds indexed by pairs $\{\vec{\beta}_{-th}, \vec{\beta}_{+th}\} = \{ \{\beta_{1,-th}, \beta_{1,+th}\}, \{\beta_{2,-th}, \beta_{2,+th}\}, \cdots, \{\beta_{m,-th}, \beta_{m,+th}\}\}$

\begin{align}
    p_{err}(\{\vec{\beta}_{-th}, \vec{\beta}_{+th}\}, \kappa) 
    = \label{eq:general-multi-thresh-perr}
    \frac{\kappa}{\pi} \sum_{j=1}^{m}(\beta_{j,+th} - \beta_{j,-th})
    +
    \sum_{s = -d}^{d} c_s 
     \left[
     \text{sinc}(\pi s)
     +
     (\frac{i}{\pi s}) \sum_{j=1}^{m} \left(
     e^{i( 2\kappa \beta_{j,+th}) s} - 
     e^{i( 2\kappa \beta_{j,-th}) s}\right)
     \right]
\end{align}

\subsection{Symmetric \texorpdfstring{$ P(M = \downarrow| \beta)$ for degree 1}{}} \label{sec:gqspi-degree1-symmetry}
The operator sequence for the degree-1 GQSPI set-up would be given by $U(\beta, \kappa):$
\begin{align*}
    \left[(R(\theta_0, -\lambda_0, -\phi_0)\otimes I_{osc})D_c(-i\kappa/\sqrt{2})\right]
    S_{\beta}
    \left[D_c(i\kappa/\sqrt{2})(R(\theta_0, \phi_0, \lambda_0)\otimes I_{osc})\right]
\end{align*}

\begin{align*}
     U(\beta, \kappa) 
     &= 
     \begin{bmatrix} 
        e^{-i(\lambda + \phi)}\cos{\theta}e^{-i \kappa \hat{x}} 
        & e^{-i\lambda}\sin{\theta}e^{i \kappa \hat{x}} \\
        e^{-i\phi}\sin{\theta}e^{-i \kappa \hat{x}} 
        & 
        -\cos{\theta}e^{i \kappa \hat{x}}
    \end{bmatrix}
    \begin{bmatrix}
        e^{i \beta \hat{p}}e^{i \kappa \hat{x}}e^{i(\lambda+ \phi)}\cos{\theta} & e^{i \beta \hat{p}}e^{i \kappa \hat{x}}e^{i\phi}\sin{\theta} \\
        e^{i \beta \hat{p}}e^{-i \kappa \hat{x}} e^{i\lambda}\sin{\theta} & 
        -e^{i \beta \hat{p}}e^{-i \kappa \hat{x}}\cos{\theta}
    \end{bmatrix}
    \\
    &= 
    \begin{bmatrix} 
    % e^{-i(\lambda + \phi)}\cos{\theta}
    e^{-i \kappa \hat{x}} e^{i \beta \hat{p}}e^{i \kappa \hat{x}}
    % e^{i(\lambda+ \phi)}\
    \cos^2{\theta} +
    % e^{-i\lambda}
    \sin^2{\theta}e^{i \kappa \hat{x}}e^{i \beta \hat{p}}e^{-i \kappa \hat{x}}
    % e^{i\lambda}\sin{\theta} 
    &
e^{-i\lambda }\cos{\theta}\sin{\theta}
[
e^{-i \kappa \hat{x}}e^{i \beta \hat{p}}e^{i \kappa \hat{x}}
- e^{i \kappa \hat{x}}e^{i \beta \hat{p}}e^{-i \kappa \hat{x}}
]
% e^{i\phi}
% - e^{-i\lambda}
% -\sin{\theta}
% \cos{\theta}
    \\
        % e^{-i\phi}
       e^{i\lambda}\cos{\theta}\sin{\theta}[ e^{-i \kappa \hat{x}}  e^{i \beta \hat{p}}e^{i \kappa \hat{x}}
        - e^{i \kappa \hat{x}} e^{i \beta \hat{p}}e^{-i \kappa \hat{x}} ]
         &
         % e^{-i\phi}
         \sin^2{\theta}e^{-i \kappa \hat{x}} e^{i \beta \hat{p}}e^{i \kappa \hat{x}}
         % e^{i\phi}
         % \sin{\theta}
        +
        \cos^2{\theta}e^{i \kappa \hat{x}}
        e^{i \beta \hat{p}}e^{-i \kappa \hat{x}}
        % \cos{\theta}
    \end{bmatrix}\\
    &= \begin{bmatrix}
    (e^{i \kappa \beta} \cos^2{\theta}
     +
    e^{-i \kappa \beta}\sin^2{\theta})e^{i \beta \hat{p}}
    &
    e^{-i\lambda }\cos{\theta}\sin{\theta}
(e^{i \kappa \beta}- e^{-i \kappa \beta})e^{i \beta \hat{p}}
\\
 e^{i\lambda}\cos{\theta}\sin{\theta}( e^{i \kappa \beta}
        - e^{-i \kappa \beta} )e^{i \beta \hat{p}}
        &
        (\sin^2{\theta} e^{i \kappa \beta} 
        +
        \cos^2{\theta}e^{-i \kappa \beta})
        e^{i \beta \hat{p}}
    \end{bmatrix}  
\end{align*}
\textbf{Input state:} $\ket{\downarrow}\ket{0}_{osc}$
\\
\textbf{Output state:} $ U(\beta, \kappa) \ket{\downarrow} \ket{0}_{osc} = 
\begin{bmatrix}
e^{i \kappa \beta} \cos^2{\theta} + e^{-i \kappa \beta}\sin^2{\theta}
    \\
    e^{i\lambda}\cos{\theta}\sin{\theta}( e^{i \kappa \beta}
        - e^{-i \kappa \beta} )
\end{bmatrix} \otimes e^{i \beta \hat{p}}\ket{0}_{osc}$

Let $\nu(\beta) = e^{i 2\kappa \beta}$,
\begin{align*}
    Prob(M = \downarrow|\beta) 
    &= \langle 0|_{osc} U(\beta, \kappa)^{\dagger}_{00}U(\beta, \kappa)_{00} |0\rangle_{osc} \\
    &= (e^{-i \kappa \beta} \cos^2{\theta} + e^{i \kappa \beta}\sin^2{\theta})(e^{i \kappa \beta} \cos^2{\theta} + e^{-i \kappa \beta}\sin^2{\theta})
    \\
    &= \cos^4{\theta}+ e^{-i 2\kappa \beta}\cos^2{\theta}\sin^2{\theta}+ e^{i 2\kappa \beta}\cos^2{\theta}\sin^2{\theta} +  \sin^4{\theta}\\
    &= (\cos^4{\theta}+\sin^4{\theta}) + \cos^2{\theta}\sin^2{\theta}\nu^{-1}(\beta) + \cos^2{\theta}\sin^2{\theta}\nu(\beta)\\
    &=c_0 + c_{-1}\nu^{-1}(\beta) + c_{1}\nu(\beta)
\end{align*}
For degree 1, $c_{-1} = c_{1}$, the response function is symmetric about $\beta=0$ and only depends on $\theta_0$ value. This behavior is also confirmed by the simulation results in Fig. \ref{fig:GQSPI-Sim-Results}.

\section{GQSPI Protocol for Stochastic \texorpdfstring{$\beta$}{}}
\subsection{Expression of \texorpdfstring{$p_{err}(\beta_{-th}, \beta_{+th}, \kappa, \sigma_{\beta}, \mu_{\beta})$}{}}
\label{sec:appndx-gaussian-beta}

Let $\beta$ follow some Gaussian distribution with mean $\mu_{\beta}$ and variance $\sigma^2_{\beta}$ given by, 
\begin{align*}
    \beta \sim \frac{1}{\sqrt{2\pi \sigma^2_{\beta}}}e^{-\frac{(\beta - \mu_{\beta})^2}{2\sigma^2_{\beta}}}
\end{align*}


Then the average decision error probability density for the protocol would be, 

\begin{align}
   & \nonumber
   p_{err}(\beta_{-th}, \beta_{+th}, \kappa, \sigma_{\beta}, \mu_{\beta})
   = \frac{\kappa}{\pi}\mathbb{E} \left[ 
    |P_{ideal} - P_{approx}(\beta) |\right  ]
   % \\
   % &= \nonumber
   % \frac{\kappa}{\pi}\mathbb{E} \left[ P_{approx}(\beta) \right]\big|_{-\frac{\pi}{2\kappa}}^{\beta_{-th}}
   % +
   %  \frac{\kappa}{\pi}\mathbb{E} \left[ P_{ideal} - P_{approx}(\beta) |\right ]\big|_{\beta_{-th}}^{\beta_{+th}}
   %  +
   %   \frac{\kappa}{\pi}\mathbb{E} \left[ P_{approx}(\beta) \right]\big|_{\beta_{+th}}^{\frac{\pi}{2\kappa}}
     \\
     &= \nonumber
     \frac{\kappa}{\pi}
     \sum_{s = -d}^{d}  \frac{c_s}{\sqrt{2\pi \sigma^2_{\beta}}} \left[
     \int_{-\frac{\pi}{2\kappa}}^{\beta_{-th}}
     e^{i(2\kappa \beta) s}
     e^{-\frac{(\beta - \mu_{\beta})^2}{2\sigma^2_{\beta}}} d\beta 
     +
     \int_{\beta_{+th}}^{\frac{\pi}{2\kappa}}
     e^{i(2\kappa \beta) s}
     e^{-\frac{(\beta - \mu_{\beta})^2}{2\sigma^2_{\beta}}} d\beta 
     \right]
     \\
      &+ \nonumber
       \frac{\kappa}{\pi}
      \int_{\beta_{-th}}^{\beta_{+th}}
     \left(1- \sum_{s = -d}^{d}  \frac{c_s}{\sqrt{2\pi \sigma^2_{\beta}}} e^{i(2\kappa \beta) s}
     e^{-\frac{(\beta - \mu_{\beta})^2}{2\sigma^2_{\beta}}}
     \right) d\beta 
     % \\
     % &= \nonumber
     %  \frac{\kappa}{\pi}
     % \sum_{s = -d}^{d}  \frac{c_s}{2} 
     % e^{-2\kappa^2 s^2 \sigma^2_{\beta}} 
     % e^{i(2\kappa s \mu_{\beta})}
     % \left[
     %  \text{erf}\left(\frac{\beta_{-th} - \mu_{\beta}}{\sqrt{2}\sigma_{\beta}} -i \sqrt{2} \kappa s \sigma_{\beta}\right)
     %  -
     %   \text{erf}\left(\frac{-\frac{\pi}{2\kappa} - \mu_{\beta}}{\sqrt{2}\sigma_{\beta}} -i \sqrt{2} \kappa s \sigma_{\beta}\right)
     %   \right]
     %   \\
     %     &+ \nonumber
     %      \frac{\kappa}{\pi}
     %     \sum_{s = -d}^{d}  \frac{c_s}{2} 
     %     e^{-2\kappa^2 s^2 \sigma^2_{\beta}} 
     %     e^{i(2\kappa s \mu_{\beta})}
     % \left[
     %   \text{erf}\left(\frac{\frac{\pi}{2\kappa} - \mu_{\beta}}{\sqrt{2}\sigma_{\beta}} -i \sqrt{2} \kappa s \sigma_{\beta}\right)
     %   -
     %   \text{erf}\left(\frac{\beta_{+th} - \mu_{\beta}}{\sqrt{2}\sigma_{\beta}} -i \sqrt{2} \kappa s \sigma_{\beta}\right)
     % \right]
     % \\
     % &+  \nonumber
     % \frac{\kappa}{\pi}
     % (\beta_{+th} - \beta_{-th})
     % - \frac{\kappa}{\pi}
     % \sum_{s = -d}^{d}  \frac{c_s}{2} 
     % e^{-2\kappa^2 s^2 \sigma^2_{\beta}} 
     % e^{i(2\kappa s \mu_{\beta})}
     % \left[
     %  \text{erf}\left(\frac{\beta_{+th} - \mu_{\beta}}{\sqrt{2}\sigma_{\beta}} -i \sqrt{2} \kappa s \sigma_{\beta}\right)
     %  -
     %   \text{erf}\left(\frac{\beta_{-th} - \mu_{\beta}}{\sqrt{2}\sigma_{\beta}} -i \sqrt{2} \kappa s \sigma_{\beta}\right)
     %   \right]
       \\
       &= \nonumber
       \frac{\kappa}{\pi}
     (\beta_{+th} - \beta_{-th})
     - \frac{\kappa}{\pi}
     \sum_{s = -d}^{d}  c_s
     e^{-2\kappa^2 s^2 \sigma^2_{\beta}} 
     e^{i(2\kappa s \mu_{\beta})}
     \left[
      \text{erf}\left(\frac{\beta_{+th} - \mu_{\beta}}{\sqrt{2}\sigma_{\beta}} -i \sqrt{2} \kappa s \sigma_{\beta}\right)
      -
       \text{erf}\left(\frac{\beta_{-th} - \mu_{\beta}}{\sqrt{2}\sigma_{\beta}} -i \sqrt{2} \kappa s \sigma_{\beta}\right)
       \right]
       \\
       &+ \label{eq:perr-beta-stochastic}
       \frac{\kappa}{\pi}
     \sum_{s = -d}^{d}  \frac{c_s}{2} 
     e^{-2\kappa^2 s^2 \sigma^2_{\beta}} 
     e^{i(2\kappa s \mu_{\beta})}
     \left[
      \text{erf}\left(\frac{\frac{\pi}{2\kappa} - \mu_{\beta}}{\sqrt{2}\sigma_{\beta}} -i \sqrt{2} \kappa s \sigma_{\beta}\right)
      -
       \text{erf}\left(\frac{-\frac{\pi}{2\kappa} - \mu_{\beta}}{\sqrt{2}\sigma_{\beta}} -i \sqrt{2} \kappa s \sigma_{\beta}\right)
       \right]
\end{align}
As seen in the previous sections, the only tweak required is updating the probability of decision error density in order to find appropriate phase angles to accommodate prior information on $\beta$.

\section{GQSPI Protocol in Presence of Dephasing Noise on Oscillator}\label{sec:appnx-gqspi-deph}

\subsection{GQSP Form for Dephasing Error}
Assume that the a dephasing error occurs after the application of each conditional displacement on the oscillator, then the GQSPI protocol gets modified in the following manner,
    \begin{align*}
    G_d(\vec{\theta}, \vec{\phi}, \lambda_0, \vec{\gamma}) 
    &= 
    \left(\prod_{i=1}^{d} (R(\theta_i, \phi_i, 0)\otimes I_{osc}) \mathcal{D}_c(i \kappa / \sqrt{2}) R_{osc}(\gamma_i) \right)(R(\theta_0, \phi_0, \lambda_0)\otimes I_{osc})
    \\
    &= \nonumber
    (R(\theta_d, \phi_d, 0)\otimes I_{osc})\mathcal{D}_c(i \kappa / \sqrt{2}) R_{osc}(\gamma_d) (R(\theta_{d-1}, \phi_{d-1}, 0)\otimes I_{osc})\mathcal{D}_c(i \kappa / \sqrt{2}) R_{osc}(\gamma_{d-1}) 
    \\
    & \nonumber
    \cdots \mathcal{D}_c(i \kappa / \sqrt{2}) R_{osc}(\gamma_{1})(R(\theta_0, \phi_0, \lambda_0)\otimes I_{osc})
    % \\
    % &= \nonumber
    % (R(\theta_d, \phi_d, 0)\otimes I_{osc})R_{osc}(\gamma_d)[R^{\dagger}_{osc}(\gamma_d)\mathcal{D}_c(i \kappa / \sqrt{2}) R_{osc}(\gamma_d)](R(\theta_{d-1}, \phi_{d-1}, 0)\otimes I_{osc})R_{osc}(\gamma_{d-1})
    % \\
    % & \nonumber
    % [R^{\dagger}_{osc}(\gamma_{d-1})\mathcal{D}_c(i \kappa / \sqrt{2}) R_{osc}(\gamma_{d-1})] \cdots
    % [R^{\dagger}_{osc}(\gamma_1)\mathcal{D}_c(i \kappa / \sqrt{2}) R_{osc}(\gamma_1)](R(\theta_0, \phi_0, \lambda_0)\otimes I_{osc})
    % \\
    % &= \nonumber
    % \prod_{i=1}^{d}R_{osc}(\gamma_i)\left(
    % \prod_{i=1}^{d}
    % (R(\theta_i, \phi_i, 0)\otimes I_{osc})\mathcal{D}_c(\alpha'_i)
    % \right)(R(\theta_0, \phi_0, \lambda_0)\otimes I_{osc}), 
    \\
    &=
    R_{osc}(\sum_{i=1}^{d}\gamma_i)\left(
    \prod_{i=1}^{d}
    (R(\theta_i, \phi_i, 0)\otimes I_{osc})\mathcal{D}_c(\alpha'_i)
    \right) (R(\theta_0, \phi_0, \lambda_0)\otimes I_{osc})
        \\
    &\text{where,}\quad \alpha'_i = \frac{i\kappa \cos\Gamma_i + \kappa \sin\Gamma_i}{\sqrt{2}} 
    % = \frac{\kappa}{\sqrt{2}}[\cos{(\frac{\pi}{2} -\Gamma_i)} + i\sin{(\frac{\pi}{2} -\Gamma_i)}] 
    = \frac{\kappa}{\sqrt{2}}e^{i(\frac{\pi}{2} -\Gamma_i)}, 
    \quad \text{and,}\quad
    \Gamma_i = \sum_{j=1}^{i} \gamma_j.
    \end{align*}
The above form can be obtained by shifting the oscillator dephasing rotation gates $R_{osc}(\gamma)$ towards the left hand side.

\subsection{Polynomial Form of GQSP with Dephasing}

Polynomial obtained for degree 1 are of the form:
\begin{align*}
     P_1(\vec{\gamma}) &= 
     e^{i(\phi_{1}+ \lambda_0)} \left(
        e^{i \phi_0} \cos{\theta_{1}} \cos{\theta_0} \mathcal{D}(\alpha'_1)
        +
       \sin{\theta_{1}}\sin{\theta_0} \mathcal{D}(-\alpha'_1)
        \right)
        \\
    Q_1(\vec{\gamma}) &= e^{i\lambda_0} \left(
        e^{i\phi_0} \sin{\theta_{1}}   \cos{\theta_0} \mathcal{D}(\alpha'_1)
        -
        \cos{\theta_{1}}\sin{\theta_0} \mathcal{D}(-\alpha'_1)
        \right)
\end{align*}

Similarly, polynomial obtained for degree 2 are:
\begin{align*}
    P_2(\vec{\gamma})  &= p_{2}
     p_{2}
     \mathcal{D}(\alpha'_{2} + \alpha'_{1})
      e^{-i\frac{\kappa^2}{2}\sin(\Gamma_2 - \Gamma_1)}
      +
      p_{-2}
     \mathcal{D}(-(\alpha'_{2} + \alpha'_{1}))
      e^{-i\frac{\kappa^2}{2}\sin(\Gamma_2 - \Gamma_1)} 
      \\
      &+
      p_{1}\mathcal{D}(\alpha'_{2} - \alpha'_{1})
      e^{i\frac{\kappa^2}{2}\sin(\Gamma_2 - \Gamma_1)}
      +
      p_{-1}\mathcal{D}(- (\alpha'_{2} - \alpha'_{1}))
      e^{i\frac{\kappa^2}{2}\sin(\Gamma_2 - \Gamma_1)} 
\end{align*}
and
\begin{align*}
    Q_2(\vec{\gamma}) &= \sin{\theta_{2}}\mathcal{D}(\alpha'_{2})P_1(\alpha_1) 
        -\cos{\theta_{2}}\mathcal{D}(-\alpha'_{2})Q_1(\alpha_1).
\end{align*}

It can be observed that the general form of the polynomial $P$ involve all possible summations (and subtractions) of the conditional displacement gates applied. This can be expressed as summation of sign bits multiplied with the displacement amounts resulting in several displacement gates with varying magnitude and direction. Therefore, the general form of the polynomial can be written as,

\begin{align}
    P_d(\vec{\gamma}) 
    &=
    \sum_{\textbf{s} \in \{-1, 1\}^d} p_{\textbf{s} } \mathcal{D}\left(
    \sum_{k=1}^{d}s_k\alpha'_k 
    \right)
    \exp{
    \left(
    -i\frac{\kappa^2}{2}
    \sum_{1 \leq l < k \leq d } s_k s_l
    \sin(\Gamma_k - \Gamma_l)
    \right)}, 
    \quad s_k \in \{-1, 1\}~\forall~k ~\in \{1,2, \cdots, d\}
\end{align}

\subsection{Overall Unitary for Dephasing GQSPI}
\begin{align}
    U(\beta, \kappa, \vec{\gamma}) 
     &= 
     \nonumber G^{-1}_{d,\vec{\theta}, \vec{\phi}, \lambda_0}(\vec{\gamma}) ~S_{\beta}~ G_{d,\vec{\theta}, \vec{\phi}, \lambda_0}(\vec{\gamma})
     \\
     &=  \nonumber
     \begin{bmatrix}
         P^{\dagger}_d(\vec{\gamma})& Q^{\dagger}_d(\vec{\gamma})
         \\
         * & *
     \end{bmatrix}
     R^{\dagger}_{osc}(\Gamma_d)
     ~S_{\beta}~
     R_{osc}(\Gamma_d)
     \begin{bmatrix}
         P_d(\vec{\gamma})& *
         \\
         Q_d(\vec{\gamma})& *
     \end{bmatrix}
     % \\
     % &= \nonumber
     % \begin{bmatrix}
     %     P^{\dagger}_d(\vec{\gamma})& Q^{\dagger}_d(\vec{\gamma})
     %     \\
     %     * & *
     % \end{bmatrix}
     % e^{i\Gamma_d \hat{n}}
     % e^{i\beta \hat{p}}
     % e^{-i\Gamma_d \hat{n}}
     % \begin{bmatrix}
     %     P_d(\vec{\gamma})& *
     %     \\
     %     Q_d(\vec{\gamma})& *
     % \end{bmatrix}
     \\
     &= \nonumber
     \begin{bmatrix}
         P^{\dagger}_d(\vec{\gamma})& Q^{\dagger}_d(\vec{\gamma})
         \\
         * & *
     \end{bmatrix}
     e^{i\beta (\hat{p}\cos{(\Gamma_d)} - \hat{x}\sin{(\Gamma_d)})}
     \begin{bmatrix}
         P_d(\vec{\gamma})& *
         \\
         Q_d(\vec{\gamma})& *
     \end{bmatrix}
     \\
     &= \nonumber
     \begin{bmatrix}
         P^{\dagger}_d(\vec{\gamma})& Q^{\dagger}_d(\vec{\gamma})
         \\
         * & *
     \end{bmatrix}
     \mathcal{D}\left( -\frac{\beta}{\sqrt{2}}e^{i\Gamma_d} \right)
     \begin{bmatrix}
         P_d(\vec{\gamma})& *
         \\
         Q_d(\vec{\gamma})& *
     \end{bmatrix}
 \end{align}

 Since,
 \begin{align*}
     e^{i\Gamma_d \hat{n}}
     e^{i\beta \hat{p}}
     e^{-i\Gamma_d \hat{n}} &= e^{i\beta \hat{p} + \sum_{j=1}^{\infty} \frac{[i\Gamma_d \hat{n}, i\beta \hat{p}]_j}{j!}} = e^{i\beta(\hat{p}\cos{(\Gamma_d)}- \hat{x}\sin{(\Gamma_d)})}
 \end{align*}

 \begin{align*}
    i\beta (\hat{p}\cos{(\Gamma_d)} - \hat{x}\sin{(\Gamma_d)})
    &= 
    i\beta [-i\frac{1}{\sqrt{2}}( a - a^{\dagger})\cos{(\Gamma_d)}
    - \frac{1}{\sqrt{2}}(a + a^{\dagger})\sin{(\Gamma_d)}]
    % \\
    % &= \beta \frac{1}{\sqrt{2}}( a - a^{\dagger})\cos{(\Gamma_d)} -i\beta\frac{1}{\sqrt{2}}(a + a^{\dagger})\sin{(\Gamma_d)}
    % \\
    % &=-\frac{1}{\sqrt{2}}\beta[\cos{(\Gamma_d)} + i\sin{(\Gamma_d)}]a^{\dagger} + \frac{1}{\sqrt{2}}\beta [\cos{(\Gamma_d)} - i\sin{(\Gamma_d)}] a
    \\
    \Rightarrow
    \beta_{disp} &= -\frac{\beta}{\sqrt{2}}[\cos{(\Gamma_d)} + i\sin{(\Gamma_d)}] = -\frac{\beta}{\sqrt{2}}e^{i\Gamma_d}
\end{align*}

In presence of dephasing, the signal $S_{\beta}$ undergoes rotation through an angle dependent on accumulated dephasing phase error.


The top-left element of the operator $U(\beta, \kappa, \vec{\gamma})_{00} $ in case of dephasing can be expressed as
\begin{align}
    U(\beta, \kappa, \vec{\gamma})_{00} 
    &=
    \nonumber
    \sum_{\textbf{s} \in \{-1, 1\}^d}  
    \sum_{\textbf{s'} \in \{-1, 1\}^d}
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    \\
    & \nonumber
    \exp{
    \left(
    i\frac{\kappa^2}{2}
    \sum_{1 \leq l < k \leq d } s_k s_l
    \sin(\Gamma_k - \Gamma_l)
    \right)}
    \exp{
    \left(
    -i\frac{\kappa^2}{2}
    \sum_{1 \leq l' < k' \leq d } s'_{k'} s'_{l'}
    \sin(\Gamma_{k'} - \Gamma_{l'})
    \right)}
    \\
    & \nonumber
    \mathcal{D}^{\dagger}\left(
    \sum_{k=1}^{d}s_k\alpha'_k 
    \right)
    \mathcal{D}\left( -\frac{\beta}{\sqrt{2}}e^{i\Gamma_d} \right)
    \mathcal{D}\left(
    \sum_{k'=1}^{d}s'_{k'}\alpha'_{k'} 
    \right)
    % \\
    % &=
    % \nonumber
    % \sum_{\textbf{s}, \textbf{s'}\in \{-1, 1\}^d}  
    % (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    % \exp{
    % \left(
    % i\frac{\kappa^2}{2}
    % \sum_{1 \leq l < k \leq d } s_k s_l
    % \sin(\Gamma_k - \Gamma_l)
    % -
    % i\frac{\kappa^2}{2}
    % \sum_{1 \leq l' < k' \leq d } s'_{k'} s'_{l'}
    % \sin(\Gamma_{k'} - \Gamma_{l'})
    % \right)}
    % \\
    % &\nonumber
    % \mathcal{D}\left(
    % -\sum_{k=1}^{d}s_k\alpha'_k 
    % \right)
    % \mathcal{D}\left( -\frac{\beta}{\sqrt{2}}e^{i\Gamma_d} \right)
    % \mathcal{D}\left(
    % \sum_{k'=1}^{d}s'_{k'}\alpha'_{k'} 
    % \right)
    \\
    &=
    \nonumber
    \sum_{\textbf{s}, \textbf{s'}\in \{-1, 1\}^d}  
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    \exp{
    \left(
    i\frac{\kappa^2}{2}
    \sum_{1 \leq l < k \leq d } s_k s_l
    \sin(\Gamma_k - \Gamma_l)
    -
    i\frac{\kappa^2}{2}
    \sum_{1 \leq l' < k' \leq d } s'_{k'} s'_{l'}
    \sin(\Gamma_{k'} - \Gamma_{l'})
    \right)}
    \\
    & \label{eq:U-deph-00}
    e^{i\kappa\beta \sum_{k=1}^{d} s_k \cos{(\Gamma_d + \Gamma_k)}}
    \mathcal{D}\left( -\frac{\beta}{\sqrt{2}}e^{i\Gamma_d} \right)
    \mathcal{D}\left(
    -\sum_{k=1}^{d}s_k\alpha'_k 
    \right)
    \mathcal{D}\left(
    \sum_{k'=1}^{d}s'_{k'}\alpha'_{k'} 
    \right)
\end{align}


Here, the displacements are swapped by the relationship,
\begin{align*}
    \mathcal{D}\left(
    -\sum_{k=1}^{d}s_k\alpha'_k 
    \right)
    \mathcal{D}\left( -\frac{\beta}{\sqrt{2}}e^{i\Gamma_d} \right)
    &=
    \mathcal{D}\left( -\frac{\beta}{\sqrt{2}}e^{i\Gamma_d} \right)
    \mathcal{D}\left(
    -\sum_{k=1}^{d}s_k\alpha'_k 
    \right)
    e^{[(\frac{\beta}{\sqrt{2}}e^{i\Gamma_d})(\sum_{k=1}^{d}s_k\alpha'_k )^* - (\frac{\beta}{\sqrt{2}}e^{-i\Gamma_d})(\sum_{k=1}^{d}s_k\alpha'_k )]}
    \\
    &=
    \mathcal{D}\left( -\frac{\beta}{\sqrt{2}}e^{i\Gamma_d} \right)
    \mathcal{D}\left(
    -\sum_{k=1}^{d}s_k\alpha'_k 
    \right)
    e^{i\kappa\beta \sum_{k=1}^{d} s_k \cos{(\Gamma_d + \Gamma_k)}}
\end{align*}

with phase calculation,
\begin{align*}
    &(\frac{\beta}{\sqrt{2}}e^{i\Gamma_d})(\sum_{k=1}^{d}s_k\alpha'_k )^* - (\frac{\beta}{\sqrt{2}}e^{-i\Gamma_d})(\sum_{k=1}^{d}s_k\alpha'_k )
    \\
    &=
    (\frac{\beta}{\sqrt{2}}e^{i\Gamma_d})(\sum_{k=1}^{d}s_k \frac{\kappa}{\sqrt{2}}e^{-i(\frac{\pi}{2} - \Gamma_k)})
    - 
    (\frac{\beta}{\sqrt{2}}e^{-i\Gamma_d})
    (\sum_{k=1}^{d}s_k \frac{\kappa}{\sqrt{2}}e^{i(\frac{\pi}{2} - \Gamma_k)})
    % \\
    % &=
    % \frac{\kappa\beta}{2}\sum_{k=1}^{d}s_k
    % \left[
    % e^{-i(\frac{\pi}{2} - \Gamma_d - \Gamma_k)}
    % -
    % e^{i(\frac{\pi}{2} - \Gamma_d - \Gamma_k)}
    % \right]
    % \\
    % &=
    %  \frac{\kappa\beta}{2} \sum_{k=1}^{d} s_k
    % \left[
    % e^{-i(\frac{\pi}{2} - \Gamma_d - \Gamma_k)}
    % -
    % e^{i(\frac{\pi}{2} - \Gamma_d - \Gamma_k)}
    % \right]
    \\
    &=
    -i\kappa\beta \sum_{k=1}^{d} s_k \sin{(\frac{\pi}{2} - \Gamma_d - \Gamma_k)}
    =
    -i\kappa\beta \sum_{k=1}^{d} s_k \cos{( - \Gamma_d - \Gamma_k)} 
    = 
    i\kappa\beta \sum_{k=1}^{d} s_k \cos{(\Gamma_d + \Gamma_k)}
\end{align*}




\subsection{\texorpdfstring{$P(M = \downarrow| \beta)$}{}}

Following a similar calculation as shown previously in Appendix~\ref{sec:prob-beta},
\begin{align*}
    P(M = \downarrow| \beta) 
    &=
    \langle 0|_{osc} U^{\dagger}(\beta, \kappa, \vec{\gamma})_{00}U(\beta, \kappa, \vec{\gamma})_{00}|0 \rangle _{osc}
    \\
    &=
    \int_{-\infty}^{\infty} 
    U^{\dagger}(\beta, \kappa, \vec{\gamma})_{00}U(\beta, \kappa, \vec{\gamma})_{00}~ \psi^{2}_0(x) dx
    \\
    &=
    \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} })
    e^{i\Theta_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}}
    e^{i\kappa\beta \Lambda_{\textbf{s}, \textbf{r}}}
   \left[
    \int_{-\infty}^{\infty} 
    \mathcal{D}(\alpha_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}})
    ~ \psi^{2}_0(x) dx
   \right]
   \\
   &=
   \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} })
    e^{i\Theta_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}}
    e^{i\kappa\beta \Lambda_{\textbf{s}, \textbf{r}}} 
    e^{-|\alpha_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}|^2/2}
\end{align*}

where,
\begin{align*}
    &U^{\dagger}(\beta, \kappa, \vec{\gamma})_{00} U(\beta, \kappa, \vec{\gamma})_{00}
    \\
    &=
    \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} })
    \exp{
    \left(
    i\frac{\kappa^2}{2}
    \sum_{1 \leq l < k \leq d } s_k s_l
    \sin(\Gamma_k - \Gamma_l)
    -
    i\frac{\kappa^2}{2}
    \sum_{1 \leq l' < k' \leq d } s'_{k'} s'_{l'}
    \sin(\Gamma_{k'} - \Gamma_{l'})
    \right)
    }
    \\
    &
    \exp{
    \left(
    -i\frac{\kappa^2}{2}
    \sum_{1 \leq n < m \leq d } s_m s_n
    \sin(\Gamma_m - \Gamma_n)
    +
    i\frac{\kappa^2}{2}
    \sum_{1 \leq n' < m' \leq d } s'_{m'} s'_{n'}
    \sin(\Gamma_{m'} - \Gamma_{n'})
    \right)
    }
    \\
    &
    e^{-i\kappa\beta \sum_{m=1}^{d} s_m \cos{(\Gamma_d + \Gamma_m)}}
    e^{i\kappa\beta \sum_{k=1}^{d} s_k \cos{(\Gamma_d + \Gamma_k)}}
    \\
    &
    \mathcal{D}{\dagger}\left(
    \sum_{m'=1}^{d}s'_{m'}\alpha'_{m'} 
    \right)
    \mathcal{D}{\dagger}\left(
    -\sum_{m=1}^{d}s_m\alpha'_m 
    \right)
    \mathcal{D}^{\dagger}\left( -\frac{\beta}{\sqrt{2}}e^{i\Gamma_d} \right)
    \mathcal{D}\left( -\frac{\beta}{\sqrt{2}}e^{i\Gamma_d} \right)
    \mathcal{D}\left(
    -\sum_{k=1}^{d}s_k\alpha'_k 
    \right)
    \mathcal{D}\left(
    \sum_{k'=1}^{d}s'_{k'}\alpha'_{k'} 
    \right)
    % \\
    % &=
    % \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    % (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    % (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} })
    % \\
    % &
    % \exp{
    % \left(
    % i\frac{\kappa^2}{2} \left[
    % \sum_{1 \leq l < k \leq d } s_k s_l
    % \sin(\Gamma_k - \Gamma_l)
    % -
    % \sum_{1 \leq l' < k' \leq d } s'_{k'} s'_{l'}
    % \sin(\Gamma_{k'} - \Gamma_{l'})
    % -
    % \sum_{1 \leq n < m \leq d } r_m r_n
    % \sin(\Gamma_m - \Gamma_n)
    %  \right. \right. }
    % \\
    % &
    % {
    % \left. \left.
    % +
    % \sum_{1 \leq n' < m' \leq d } r'_{m'} r'_{n'}
    % \sin(\Gamma_{m'} - \Gamma_{n'})
    % \right]
    % \right)
    % }
    % e^{i\kappa\beta [\sum_{k=1}^{d} s_k\cos{(\Gamma_d + \Gamma_k)} - \sum_{m=1}^{d}r_m\cos{(\Gamma_d + \Gamma_m)}]}
    % \quad \quad \quad
    % (\text{call it }e^{i\Phi_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}}, e^{i\kappa\beta \Lambda_{\textbf{s}, \textbf{r}}})
    % \\
    % &
    % \mathcal{D}\left(
    % -\sum_{m'=1}^{d}r'_{m'}\alpha'_{m'} 
    % \right)
    % \mathcal{D}\left(
    % \sum_{m=1}^{d}r_m\alpha'_m 
    % \right)
    % \mathcal{D}\left(
    % -\sum_{k=1}^{d}s_k\alpha'_k 
    % \right)
    % \mathcal{D}\left(
    % \sum_{k'=1}^{d}s'_{k'}\alpha'_{k'} 
    % \right)
    \\
    &=
    \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} })
    e^{i\Phi_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}}e^{i\Phi'_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}}
    e^{i\kappa\beta \Lambda_{\textbf{s}, \textbf{r}}}
    \mathcal{D}(\alpha_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}})
    \\
    &=
    \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} })
    e^{i\Theta_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}}
    e^{i\kappa\beta \Lambda_{\textbf{s}, \textbf{r}}} 
    \mathcal{D}(\alpha_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}) 
    \quad\quad 
    \text{where, }
    (\Theta_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}} = \Phi_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}+\Phi'_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}})
\end{align*}

The expressions for the factors are:
\begin{align}
    e^{i\Phi_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}}
    &= \nonumber
    \exp{
    \left(
    i\frac{\kappa^2}{2} \left[
    \sum_{1 \leq l < k \leq d } s_k s_l
    \sin(\Gamma_k - \Gamma_l)
    -
    \sum_{1 \leq l' < k' \leq d } s'_{k'} s'_{l'}
    \sin(\Gamma_{k'} - \Gamma_{l'})
    -
    \sum_{1 \leq n < m \leq d } r_m r_n
    \sin(\Gamma_m - \Gamma_n)
     \right. \right. }
    \\
    &{
    \left. \left.
    +
    \sum_{1 \leq n' < m' \leq d } r'_{m'} r'_{n'}
    \sin(\Gamma_{m'} - \Gamma_{n'})
    \right]
    \right)
    }
\end{align}
\begin{align}
    e^{i\Phi'_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}}
    &=  \nonumber
    \exp{\left( 
    \sum_{m',m = 1}^{d} ir'_{m'}r_m\frac{\kappa^2}{2}\sin{(\Gamma_{m'} - \Gamma_{m})}
    -
    \sum_{m', k = 1}^{d} ir'_{m'}s_k\frac{\kappa^2}{2}\sin{(\Gamma_{m'} - \Gamma_{k})}
    +
    \sum_{m', k' = 1}^{d} ir'_{m'}s_{k'}\frac{\kappa^2}{2}\sin{(\Gamma_{m'} - \Gamma_{k'})}
    \right. 
    }
    \\
    &{
    \left.
    + \sum_{m, k = 1}^{d} ir_{m}s_k\frac{\kappa^2}{2}\sin{(\Gamma_{m} - \Gamma_{k})}
    -
    \sum_{m, k' = 1}^{d} ir_{m}s'_{k'}\frac{\kappa^2}{2}\sin{(\Gamma_{m} - \Gamma_{k'})}
    +
    \sum_{k,k' = 1}^{d} is_{k}s'_{k'}\frac{\kappa^2}{2}\sin{(\Gamma_{k} - \Gamma_{k'})}
    \right)} 
\end{align}

\begin{align}
    e^{i\kappa\beta \Lambda_{\textbf{s}, \textbf{r}}} 
    =
    e^{i\kappa\beta [\sum_{k=1}^{d} s_k\cos{(\Gamma_d + \Gamma_k)} - \sum_{m=1}^{d}r_m\cos{(\Gamma_d + \Gamma_m)}]}
\end{align}

\begin{align}
    \mathcal{D}(\alpha_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}})
    &=
    \mathcal{D}\left(
    -\sum_{m'=1}^{d}r'_{m'}\alpha'_{m'} 
    +
    \sum_{m=1}^{d}r_m\alpha'_m 
    -
    \sum_{k=1}^{d}s_k\alpha'_k 
    +
    \sum_{k'=1}^{d}s'_{k'}\alpha'_{k'} 
    \right)
\end{align}

These factors would simplify further as many of the arguments either add up or cancel out based on the composition of the displacement gates.
\subsection{Effect of small \texorpdfstring{$\gamma$}{} on \texorpdfstring{$P(M = \downarrow| \beta)$}{}}


For small $\gamma$, we can approximate 
\begin{align*}
    sin{(a\gamma)} \approx a\gamma - \frac{a^3\gamma^3}{6} + O(\gamma^5), \quad \cos{(a\gamma)} \approx  1 - \frac{a^2\gamma^2}{2} + O(\gamma^4)  
\end{align*}

We first approximate each factor in $P(M = \downarrow| \beta)$. 

On computing the argument of $e^{i\Theta_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}}$, it can be observed that it would consist of a summation of sinusoids with some coefficients. Thus,
\begin{align}
    e^{i\Theta_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}
     }
     & \nonumber
     \equiv e^{i\frac{\kappa^2}{2}\sum_a A_a \sin{a\gamma}}, \quad a \in \mathbb{Z}, A_a \in \mathbb{R}
     \\ 
     &\approx e^{i\frac{\kappa^2}{2} (A \gamma+ A' \gamma^3)}, \quad A, A' \in \mathbb{R}
\end{align}

Similarly, since $s_k$ and $r_m$ are both sign values $\pm 1$, the cosines with same argument either add up or cancel out.
\begin{align}
    e^{i\kappa\beta \Lambda_{\textbf{s}, \textbf{r}}} 
    &= \nonumber
    e^{i\kappa\beta [\sum_{k=1}^{d} s_k\cos{(\Gamma_d + \Gamma_k)} - \sum_{m=1}^{d}r_m\cos{(\Gamma_d + \Gamma_m)}]}
    \equiv e^{i2\kappa\beta \sum_b \cos{(b\gamma})}, \quad b \in \mathbb{Z}
    \\
    \nonumber
    &
    \approx e^{i2\kappa\beta B}e^{-i2\kappa\beta B'\gamma^2}
    \quad B, B' \in \mathbb{R}
\end{align}


\begin{align*}
    \mathcal{D}(\alpha_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}})
    &\approx 
    \mathcal{D}\left(\sum_c C_c \frac{\kappa}{\sqrt{2}} e^{i(\frac{\pi}{2} - c\gamma)}\right)
    \approx \mathcal{D}
    \left(
    \sum_c C_c \frac{\kappa}{\sqrt{2}} 
    (i\cos{c\gamma} + \sin{c\gamma})
    \right),\quad  C_c \in \mathbb{R}
    \\
    % &\approx \mathcal{D}
    % \left(
    % \sum_c \frac{\kappa}{\sqrt{2}} 
    % (i(1 - c^2\gamma^2) + c\gamma)
    % \right)
    &\approx
    \mathcal{D}
    \left(
     \frac{\kappa}{\sqrt{2}} 
    [i(C - C'\gamma^2) + C''\gamma]
    \right), \quad\quad C,C',C'' \in \mathbb{R}
    \\
  e^{-|\alpha_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}|^2/2}  
  &\equiv e^{-(D + D'\gamma^4 + D''\gamma^2)/2}\quad\quad D, D', D'' \in \mathbb{R}
\end{align*}


\begin{align*}
    &P(M = \downarrow| \beta) 
   =
   \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} })
    e^{i(\Theta_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}
     + \Phi_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}})}
    e^{i\kappa\beta \Lambda_{\textbf{s}, \textbf{r}}} 
    e^{-|\alpha_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}}|^2/2}
    \\
    &=
    \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} })
    e^{i\frac{\kappa^2}{2} (A \gamma+ A' \gamma^3)}
    e^{i2\kappa\beta B}e^{-i2\kappa\beta B'\gamma^2}
    e^{-\kappa^2(D + D'\gamma^4 + D''\gamma^2)/2}
    \\
    &\approx
    \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} })
    [1+i\frac{\kappa^2}{2} (A \gamma+ A' \gamma^3) - \frac{\kappa^4}{4} (A \gamma+ A' \gamma^3)^2]
    \\
    &
    (1-i2\kappa\beta B'\gamma^2 - 2\kappa^2\beta^2 B'^2\gamma^4)
    [1-\frac{\kappa^2}{2}(D'\gamma^4 + D''\gamma^2)]e^{i2\kappa\beta B}e^{-D\frac{\kappa^2}{2}}
    % \\
    % &\approx
    % \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    % (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    % (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} })
    % [1+i\frac{\kappa^2}{2} (A \gamma+ A' \gamma^3) - \frac{\kappa^4}{8} (A^2 \gamma^2+ O(\gamma^4))]
    % \\
    % &
    % (1-i2\kappa\beta B\gamma^2 - 2\kappa^2\beta^2 B^2\gamma^4)
    % (1-\frac{\kappa^2}{8}(G^2\gamma^4 + G''\gamma^2))e^{i2\kappa\beta C}e^{V\frac{\kappa^2}{2}}
    \\
    &\approx
    \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} })
    [1+iA_1 \gamma + A_2\gamma^2 +iA_3 \gamma^3 + O(\gamma^4)](1-iB_1\beta\gamma^2 - O(\gamma^4)), 
    \quad A_1, A_2, A_3, B_1, C_1 \in \mathbb{R}
    \\
    &
    (1- C_1\gamma^2 - O(\gamma^4))e^{i2\kappa\beta B}e^{-D\frac{\kappa^2}{2}}
    \\
    &\approx
     \sum_{\textbf{s}, \textbf{s'}, \textbf{r}, \textbf{r'}\in \{-1, 1\}^d}  
    (p^*_{\textbf{s}}p_{\textbf{s'} } + q^*_{\textbf{s} }q_{\textbf{s'} })
    (p_{\textbf{r}}p^*_{\textbf{r'} } + q_{\textbf{r} }q^*_{\textbf{r'} }) 
    e^{i2\kappa\beta B}
    e^{-D\frac{\kappa^2}{2}}
    [1 - \Omega_1 \gamma^2 - \Omega_2\beta \gamma^3 + i(\Omega_3 \gamma + \Omega_4 \gamma^3)+\mathcal{O}(\gamma^4)],
    \quad \Omega_1, \Omega_2\in\mathbb{R}
\end{align*}

Since the probability is real-valued, the linear imaginary term must cancel. Consequently, the probability response function exhibits damping due to the second- and third-order terms in $\gamma$, along with slight deviations arising from the $\beta$-dependent third-order term as $\beta$ moves further away from $\beta = 0$ and the threshold values.

% \end{appendices}