Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Auditing Random Tensor Networks: A Quantitative Anatomy of Bond-Dimension Geometry

T. Weinmann

Published Nov 3, 2026• 0 citations
Publisher

Abstract

The random-tensor-network route to holographic geometry reads a spacetime out of a graph whose edges carry bond dimensions: entanglement entropies are minimal cuts at leading order, the cut weights are the geometric areas, and the boundary theory inherits subregion duality and error correction. This paper is an audit of that route, in the spirit of the companion audits of entanglement equilibrium and of mutual-information metric reconstruction: every assumption beyond established mathematics is isolated as one of six explicit hypotheses T1 to T6 and each is proved, checked numerically, or recorded as input. The audit yields: (i) on trees the random-tensor entropy is exactly Page, the quantum correction to the geometric area law is bond-dimension independent and the geometric limit does not remove it, verified by Monte Carlo to 10^-4; (ii) the continuum limit of the min-cut law is a property of the region-graph alignment, not of the geometry, showing that the quasi-uniform families that carried previous audits do not support the random-tensor continuum limit at all; (iii) with uniform bond dimensions the boundary entropy scaling is geometry-blind, meaning the celebrated holographic profiles must be put into the bond-dimension function by hand and geometry is strictly an input of the model; (iv) single-boundary-site mutual information is a finite-size effect and vanishes with system size, showing that the random-tensor route's geometry lives in region-level cut homology in sharp contrast to single-site mutual information routes. The ledger records the status of every hypothesis.

DOI
10.5281/zenodo.23928213

Related papers

Auditing Mutual Information: A Quantitative Anatomy of Metric Reconstruction from Entanglement

T. Weinmann

The mutual-information route to emergent geometry reads a Riemannian metric off the entanglement of a network state: lengths from a logarithmic functional of pairwise mutual information, distances from shortest paths. This paper is an audit of that route, in the spirit of the companion audit of entanglement equilibrium: every assumption beyond established mathematics is isolated as one of six explicit hypotheses M1 to M6, and each is proved, checked numerically, or refuted on the same exactly solvable pair (two-qubit) sector, supplemented by two synthetic models (a power-law chain, and a quasi-uniform versus designed point-set comparison on S2). The audit yields: M1, the exponential decay law I ~ I* exp(-d/xi), is not a technicality but the load-bearing assumption: it is equivalent to additivity of the length functional, and for power-law mutual information the reconstruction on the complete edge set produces a logarithmic remetrization d_k(u,v) = a + b ln d_h(u,v), verified by exact all-pairs shortest paths - a qualitative failure that no conformal rescaling repairs. Metric convergence (M2), measured by exact Dijkstra reconstruction on S2, holds with rate sup(d_k - d_h) <= 7.3 x 10^-3 at N = 3600 for quasi-random (golden-spiral) sets, while a designed banded equal-area set is disconnected at N <= 900 (isolated polar bands) and inflates the error by a factor of approximately 12 - local isotropy, not density uniformity, controls reconstruction. The area law with finite G (M4) carries a systematic deficit epsilon in [-0.16, -0.02] on quasi-random sets, consistent with the quasi-uniform bound xi + delta/xi. A pre-registered control protocol attributing the deficit to density fluctuations is executed and refuted (the registered correlation has the wrong sign, r = -0.37), locating the leading systematic in the boundary discretization of the regions, exactly the alternative pre-registered in the companion paper. Curvature extraction (M5) is one-sidedly biased (d_k >= d_h always) and only marginally resolvable at accessible sizes. The audit locates the quantitative obstructions precisely: in the decay law (M1), in the edge convention, and in the boundary layer. The ledger records the status of every hypothesis.

•Nov 3, 2026•0 citations

Entanglement Before Geometry: Random Tensor Networks as the Quantum Completion of the Flux-Gauge Area Law

Mohammad Hannan

Abstract A confining, non-Abelian gauge connection Uij ∈ SU(N) defined on the links of a dynamical graph has been proposed as the microscopic origin of classical spacetime geometry: the metric is reconstructed from the coset (soldering-form) sector of an enlarged connection, and black-hole and boundary-cut entropy are obtained as the logarithm of a combinatorial microstate count, matched, via the Freedman–Headrick theorem, to the bit-thread (max-flow/min-cut) formulation of holographic entanglement entropy. That matching is exact as a statement about counting, but it is not, by itself, a statement about the entanglement entropy of any specific quantum state: nothing in the combinatorial construction distinguishes a classical ensemble of flux occupation numbers from a genuine quantum wavefunction, so its connection to the entanglement/tensornetwork program of emergent geometry (Van Raamsdonk, Swingle, Ryu–Takayanagi, and the random-tensor-network derivation of Hayden, Qi, Kim, and Preskill) remains an analogy rather than a derivation. This paper closes that gap. We first restate the flux-gauge connection ensemble in full – its partition function, its block-link renormalization-group coarse-graining with closed-form Migdal–Kadanoff coefficients, its emergent metric and Newton’s-constant identification, and its combinatorial boundary-cut entropy – so that the paper is self-contained. We then show, by an exact character-expansion/Peter–Weyl derivation standard in lattice gauge theory but not previously applied to this construction, that the SU(N) link ensemble is a tensor network: each flux-tube segment (link) carries a bond dimension equal to the dimension of the SU(N) representation it transports, and each flux-tube junction (node) carries a gauge-invariant intertwiner tensor, fixed by representation theory rather than assumed. Only a single, narrowly scoped statistical assumption – on the distribution of intertwiners within the multiplicity space that gauge invariance already guarantees – is needed to complete this exact, deterministic tensor network into a genuine random tensor network in the technical sense of Hayden–Qi–Kim–Preskill (HQKP). We further identify the renormalization-group layers of Section 2’s coarse-graining as the isometric maps of a multi-scale entanglement renormalization ansatz (MERA), giving the emergent radial holographic direction a literal tensor-network origin rather than a scaling-dimension analogy. We then investigate, analytically and how the entanglement entropy of such a network reproduces the Ryu–Takayanagi/min-cut law. An exactly solvable limit built from classical (GHZ-type) perfect tensors reproduces the combinatorial law exactly for single-cut (boundary-anchored) regions but is shown to be degenerate for interior, two-sided regions, exposing a genuine gap in the combinatorial entropy argument: classical flux-occupation counting does not, by itself, certify multi-region (entanglement-wedgelevel) quantum entanglement structure. We then construct a genuinely quantum (non-classical, Gaussian-random) tensor-network realization of the link ensemble and confirm numerically, on chains of up to sixteen sites, (i) a clean area-law plateau in which the entanglement entropy of an interior region becomes independent of its size once the region is large enough, in place of continued volume-law growth, and (ii) a linear scaling of the plateau entropy with the logarithm of the bond dimension, S(A) = κNcut(A) logD, fit with R2 = 0.996 and an O(1) efficiency coefficient κ ≈ 0.68 for this finite, non-isometric network. We interpret κ < 1 as a genuine, quantifiable finite-network correction relative to the idealized D → ∞ isometric-projector limit in which HQKP prove κ → 1, and we list the construction of that idealized limit, directly on the SU(N) link ensemble, as the key open numerical task. We close by proposing an explicit correspondence between the tensor-network minimal surface of this paper and the Cartan soldering-form metric of the gauge-connection construction, framed as a falsifiable consistency conjecture rather than an established result.

•Jul 20, 2026•0 citations

Auditing Holographic Entanglement Entropy: A Quantitative Anatomy of the Ryu-Takayanagi Prescription

T. Weinmann

I subject the Ryu-Takayanagi (RT) prescription for holographic entanglement entropy to the same hypothesis-led audit previously applied to entanglement equilibrium, mutual information reconstruction, and random tensor networks. The prescription is decomposed into six falsifiable hypotheses T1 to T6. The main findings are: On the exactly solvable AdS3/CFT2 Poincare disk we verify the geodesic-length formula, its logarithmic coefficient (Brown-Henneaux matching), and characterize the mutual-information phase transition quantitatively. The transition sits at cross ratio xc = 1/2 + O(exp(-2rho)), exactly at alpha_c = pi/2 for all rho in the antipodal-symmetric family, with randomized-asymmetric deviations vanishing exponentially with cutoff depth rho, showing that the universal CFT2 location is the asymptotic limit of the bulk criterion, not an identity at finite cutoff. The homology constraint, often presented as an output of the holographic setup, is shown to be an input: it must be imposed by hand on the graph, and the phase transition it produces is first-order-sharp at the RT level while any discretization smears it. The continuum limit of discretized RT surfaces fails to be Riemannian: polar-lattice min-cuts converge, under delta-refinement, to a Finsler (taxicab-normal) metric with a systematic, delta-independent offset of +4.5% at 60 degrees, +13% at 30 degrees, +18% at 15 degrees, and -0.1% at 90 degrees. Mutual information, a difference of areas, amplifies these offsets asymmetrically: the connected-phase signal is underreported by up to 24% at delta = 0.1, while the transition window fabricates spurious mutual information of +0.39 at delta = 0.2. I conclude that RT is exact as a formula, input-carrying as a prescription, and metric-blind under discretization, with the error concentrated precisely at the shallow geodesics where the prescription is most often probed. All numbers are generated by two deterministic Python programs with no Monte Carlo required.

•Nov 3, 2026•0 citations

Holographic duality from random tensor networks

Patrick M. Hayden, Sepehr Nezami, Xiao-Liang Qi +3 more

Tensor networks provide a natural framework for exploring holographic duality because they obey entanglement area laws. They have been used to construct explicit toy models realizing many of the interesting structural features of the AdS/CFT correspondence, including the non-uniqueness of bulk operator reconstruction in the boundary theory. In this article, we explore the holographic properties of networks of random tensors. We find that our models naturally incorporate many features that are analogous to those of the AdS/CFT correspondence. When the bond dimension of the tensors is large, we show that the entanglement entropy of all boundary regions, whether connected or not, obey the Ryu-Takayanagi entropy formula, a fact closely related to known properties of the multipartite entanglement of assistance. We also discuss the behavior of Rényi entropies in our models and contrast it with AdS/CFT. Moreover, we find that each boundary region faithfully encodes the physics of the entire bulk entanglement wedge, i.e., the bulk region enclosed by the boundary region and the minimal surface. Our method is to interpret the average over random tensors as the partition function of a classical ferromagnetic Ising model, so that the minimal surfaces of Ryu-Takayanagi appear as domain walls. Upon including the analog of a bulk field, we find that our model reproduces the expected corrections to the Ryu-Takayanagi formula: the bulk minimal surface is displaced and the entropy is augmented by the entanglement of the bulk field. Increasing the entanglement of the bulk field ultimately changes the minimal surface behavior topologically, in a way similar to the effect of creating a black hole. Extrapolating bulk correlation functions to the boundary permits the calculation of the scaling dimensions of boundary operators, which exhibit a large gap between a small number of low-dimension operators and the rest. While we are primarily motivated by the AdS/CFT duality, the main results of the article define a more general form of bulk-boundary correspondence which could be useful for extending holography to other spacetimes.

arXiv:1601.01694•Journal of High Energy Physics•Nov 1, 2016•687 citations

Superposed Random Spin Tensor Networks and their Holographic Properties

Simon Langenscheidt

We study criteria for and properties of boundary-to-boundary holography in a class of spin network states defined by analogy to projected entangled pair states (PEPS). In particular, we consider superpositions of states corresponding to well-defined, discrete geometries on a graph. By applying random tensor averaging techniques, we map entropy calculations to a random Ising model on the same graph, with distribution of couplings determined by the relative sizes of the involved geometries. The superposition of tensor network states with variable bond dimension used here presents a picture of a genuine quantum sum over geometric backgrounds. We find that, whenever each individual geometry produces an isometric mapping of a fixed boundary region C to its complement, then their superposition does so iff the relative weight going into each geometry is inversely proportional to its size. Additionally, we calculate average and variance of the area of the given boundary region and find that the average is bounded from below and above by the mean and sum of the individual areas, respectively. Finally, we give an outlook on possible extensions to our program and highlight conceptual limitations to implementing these.

arXiv:2205.09761•May 19, 2022•0 citations

Boundary mutual information in double holography

Yuxuan Liu, Yi Ling, Zhuo-Yu Xian

We consider a composite system where AdS$_3$ gravity is coupled to a flat heat bath and investigate the mutual information between two subregions on the intersection of the AdS$_3$ and bath, referred to as the boundary mutual information (BMI). The corresponding entanglement entropy is captured via quantum extremal surfaces (QES), which holographically be computed by a surface optimization algorithm based on ``Surface Evolver''. We focus on both connected and disconnected configurations of the quantum entanglement wedge (Q-EW) in the AdS$_3$ bulk and analyze the finite corrections to the BMI. Our numerical results reveal a phase transition of the BMI as the separation between two subregions increases. Furthermore, we find that the BMI can naturally be decomposed into two distinct components: a geometric term arising from the areas of the quantum extremal surfaces, and a correction term resulting from bulk quantum fields within the Q-EW. Interestingly, the geometric contribution always exceeds the total BMI, indicating a negative correction from the bulk matter fields. This negativity can be understood as the result of subtracting a greater contribution from quantum fields in the connected Q-EW than in the disconnected one. We also reproduce the negative contribution of bulk quantum fields to BMI within a random tensor network (RTN) toy model of double holography. Modeling the bulk as a highly mixed state entangled with a large bath leads to a volume-law bulk entropy. In the large bond-dimension limit, the geometric part of the BMI remains non-negative, while the bulk entropy contribution becomes non-positive when the Q-EWs merge.

hep-thCross-listedarXiv:2602.12627•Journal of High Energy Physics•Feb 16, 2026•0 citations
The random-tensor-network route to holographic geometry reads a spacetime out of a graph whose edges carry bond dimensions: entanglement entropies are minimal cuts at leading order, the cut weights are the geometric areas, and the boundary theory inherits subregion duality and error correction. This paper is an audit of that route, in the spirit of the companion audits of entanglement equilibrium and of mutual-information metric reconstruction: every assumption beyond established mathematics is isolated as one of six explicit hypotheses T1 to T6 and each is proved, checked numerically, or recorded as input. The audit yields: (i) on trees the random-tensor entropy is exactly Page, the quantum correction to the geometric area law is bond-dimension independent and the geometric limit does not remove it, verified by Monte Carlo to 10^-4; (ii) the continuum limit of the min-cut law is a property of the region-graph alignment, not of the geometry, showing that the quasi-uniform families that carried previous audits do not support the random-tensor continuum limit at all; (iii) with uniform bond dimensions the boundary entropy scaling is geometry-blind, meaning the celebrated holographic profiles must be put into the bond-dimension function by hand and geometry is strictly an input of the model; (iv) single-boundary-site mutual information is a finite-size effect and vanishes with system size, showing that the random-tensor route's geometry lives in region-level cut homology in sharp contrast to single-site mutual information routes. The ledger records the status of every hypothesis.
The mutual-information route to emergent geometry reads a Riemannian metric off the entanglement of a network state: lengths from a logarithmic functional of pairwise mutual information, distances from shortest paths. This paper is an audit of that route, in the spirit of the companion audit of entanglement equilibrium: every assumption beyond established mathematics is isolated as one of six explicit hypotheses M1 to M6, and each is proved, checked numerically, or refuted on the same exactly solvable pair (two-qubit) sector, supplemented by two synthetic models (a power-law chain, and a quasi-uniform versus designed point-set comparison on S2). The audit yields: M1, the exponential decay law I ~ I* exp(-d/xi), is not a technicality but the load-bearing assumption: it is equivalent to additivity of the length functional, and for power-law mutual information the reconstruction on the complete edge set produces a logarithmic remetrization d_k(u,v) = a + b ln d_h(u,v), verified by exact all-pairs shortest paths - a qualitative failure that no conformal rescaling repairs. Metric convergence (M2), measured by exact Dijkstra reconstruction on S2, holds with rate sup(d_k - d_h) <= 7.3 x 10^-3 at N = 3600 for quasi-random (golden-spiral) sets, while a designed banded equal-area set is disconnected at N <= 900 (isolated polar bands) and inflates the error by a factor of approximately 12 - local isotropy, not density uniformity, controls reconstruction. The area law with finite G (M4) carries a systematic deficit epsilon in [-0.16, -0.02] on quasi-random sets, consistent with the quasi-uniform bound xi + delta/xi. A pre-registered control protocol attributing the deficit to density fluctuations is executed and refuted (the registered correlation has the wrong sign, r = -0.37), locating the leading systematic in the boundary discretization of the regions, exactly the alternative pre-registered in the companion paper. Curvature extraction (M5) is one-sidedly biased (d_k >= d_h always) and only marginally resolvable at accessible sizes. The audit locates the quantitative obstructions precisely: in the decay law (M1), in the edge convention, and in the boundary layer. The ledger records the status of every hypothesis.
Abstract A confining, non-Abelian gauge connection Uij ∈ SU(N) defined on the links of a dynamical graph has been proposed as the microscopic origin of classical spacetime geometry: the metric is reconstructed from the coset (soldering-form) sector of an enlarged connection, and black-hole and boundary-cut entropy are obtained as the logarithm of a combinatorial microstate count, matched, via the Freedman–Headrick theorem, to the bit-thread (max-flow/min-cut) formulation of holographic entanglement entropy. That matching is exact as a statement about counting, but it is not, by itself, a statement about the entanglement entropy of any specific quantum state: nothing in the combinatorial construction distinguishes a classical ensemble of flux occupation numbers from a genuine quantum wavefunction, so its connection to the entanglement/tensornetwork program of emergent geometry (Van Raamsdonk, Swingle, Ryu–Takayanagi, and the random-tensor-network derivation of Hayden, Qi, Kim, and Preskill) remains an analogy rather than a derivation. This paper closes that gap. We first restate the flux-gauge connection ensemble in full – its partition function, its block-link renormalization-group coarse-graining with closed-form Migdal–Kadanoff coefficients, its emergent metric and Newton’s-constant identification, and its combinatorial boundary-cut entropy – so that the paper is self-contained. We then show, by an exact character-expansion/Peter–Weyl derivation standard in lattice gauge theory but not previously applied to this construction, that the SU(N) link ensemble is a tensor network: each flux-tube segment (link) carries a bond dimension equal to the dimension of the SU(N) representation it transports, and each flux-tube junction (node) carries a gauge-invariant intertwiner tensor, fixed by representation theory rather than assumed. Only a single, narrowly scoped statistical assumption – on the distribution of intertwiners within the multiplicity space that gauge invariance already guarantees – is needed to complete this exact, deterministic tensor network into a genuine random tensor network in the technical sense of Hayden–Qi–Kim–Preskill (HQKP). We further identify the renormalization-group layers of Section 2’s coarse-graining as the isometric maps of a multi-scale entanglement renormalization ansatz (MERA), giving the emergent radial holographic direction a literal tensor-network origin rather than a scaling-dimension analogy. We then investigate, analytically and how the entanglement entropy of such a network reproduces the Ryu–Takayanagi/min-cut law. An exactly solvable limit built from classical (GHZ-type) perfect tensors reproduces the combinatorial law exactly for single-cut (boundary-anchored) regions but is shown to be degenerate for interior, two-sided regions, exposing a genuine gap in the combinatorial entropy argument: classical flux-occupation counting does not, by itself, certify multi-region (entanglement-wedgelevel) quantum entanglement structure. We then construct a genuinely quantum (non-classical, Gaussian-random) tensor-network realization of the link ensemble and confirm numerically, on chains of up to sixteen sites, (i) a clean area-law plateau in which the entanglement entropy of an interior region becomes independent of its size once the region is large enough, in place of continued volume-law growth, and (ii) a linear scaling of the plateau entropy with the logarithm of the bond dimension, S(A) = κNcut(A) logD, fit with R2 = 0.996 and an O(1) efficiency coefficient κ ≈ 0.68 for this finite, non-isometric network. We interpret κ < 1 as a genuine, quantifiable finite-network correction relative to the idealized D → ∞ isometric-projector limit in which HQKP prove κ → 1, and we list the construction of that idealized limit, directly on the SU(N) link ensemble, as the key open numerical task. We close by proposing an explicit correspondence between the tensor-network minimal surface of this paper and the Cartan soldering-form metric of the gauge-connection construction, framed as a falsifiable consistency conjecture rather than an established result.
I subject the Ryu-Takayanagi (RT) prescription for holographic entanglement entropy to the same hypothesis-led audit previously applied to entanglement equilibrium, mutual information reconstruction, and random tensor networks. The prescription is decomposed into six falsifiable hypotheses T1 to T6. The main findings are: On the exactly solvable AdS3/CFT2 Poincare disk we verify the geodesic-length formula, its logarithmic coefficient (Brown-Henneaux matching), and characterize the mutual-information phase transition quantitatively. The transition sits at cross ratio xc = 1/2 + O(exp(-2rho)), exactly at alpha_c = pi/2 for all rho in the antipodal-symmetric family, with randomized-asymmetric deviations vanishing exponentially with cutoff depth rho, showing that the universal CFT2 location is the asymptotic limit of the bulk criterion, not an identity at finite cutoff. The homology constraint, often presented as an output of the holographic setup, is shown to be an input: it must be imposed by hand on the graph, and the phase transition it produces is first-order-sharp at the RT level while any discretization smears it. The continuum limit of discretized RT surfaces fails to be Riemannian: polar-lattice min-cuts converge, under delta-refinement, to a Finsler (taxicab-normal) metric with a systematic, delta-independent offset of +4.5% at 60 degrees, +13% at 30 degrees, +18% at 15 degrees, and -0.1% at 90 degrees. Mutual information, a difference of areas, amplifies these offsets asymmetrically: the connected-phase signal is underreported by up to 24% at delta = 0.1, while the transition window fabricates spurious mutual information of +0.39 at delta = 0.2. I conclude that RT is exact as a formula, input-carrying as a prescription, and metric-blind under discretization, with the error concentrated precisely at the shallow geodesics where the prescription is most often probed. All numbers are generated by two deterministic Python programs with no Monte Carlo required.
Tensor networks provide a natural framework for exploring holographic duality because they obey entanglement area laws. They have been used to construct explicit toy models realizing many of the interesting structural features of the AdS/CFT correspondence, including the non-uniqueness of bulk operator reconstruction in the boundary theory. In this article, we explore the holographic properties of networks of random tensors. We find that our models naturally incorporate many features that are analogous to those of the AdS/CFT correspondence. When the bond dimension of the tensors is large, we show that the entanglement entropy of all boundary regions, whether connected or not, obey the Ryu-Takayanagi entropy formula, a fact closely related to known properties of the multipartite entanglement of assistance. We also discuss the behavior of Rényi entropies in our models and contrast it with AdS/CFT. Moreover, we find that each boundary region faithfully encodes the physics of the entire bulk entanglement wedge, i.e., the bulk region enclosed by the boundary region and the minimal surface. Our method is to interpret the average over random tensors as the partition function of a classical ferromagnetic Ising model, so that the minimal surfaces of Ryu-Takayanagi appear as domain walls. Upon including the analog of a bulk field, we find that our model reproduces the expected corrections to the Ryu-Takayanagi formula: the bulk minimal surface is displaced and the entropy is augmented by the entanglement of the bulk field. Increasing the entanglement of the bulk field ultimately changes the minimal surface behavior topologically, in a way similar to the effect of creating a black hole. Extrapolating bulk correlation functions to the boundary permits the calculation of the scaling dimensions of boundary operators, which exhibit a large gap between a small number of low-dimension operators and the rest. While we are primarily motivated by the AdS/CFT duality, the main results of the article define a more general form of bulk-boundary correspondence which could be useful for extending holography to other spacetimes.
We study criteria for and properties of boundary-to-boundary holography in a class of spin network states defined by analogy to projected entangled pair states (PEPS). In particular, we consider superpositions of states corresponding to well-defined, discrete geometries on a graph. By applying random tensor averaging techniques, we map entropy calculations to a random Ising model on the same graph, with distribution of couplings determined by the relative sizes of the involved geometries. The superposition of tensor network states with variable bond dimension used here presents a picture of a genuine quantum sum over geometric backgrounds. We find that, whenever each individual geometry produces an isometric mapping of a fixed boundary region C to its complement, then their superposition does so iff the relative weight going into each geometry is inversely proportional to its size. Additionally, we calculate average and variance of the area of the given boundary region and find that the average is bounded from below and above by the mean and sum of the individual areas, respectively. Finally, we give an outlook on possible extensions to our program and highlight conceptual limitations to implementing these.
We consider a composite system where AdS3_33​ gravity is coupled to a flat heat bath and investigate the mutual information between two subregions on the intersection of the AdS3_33​ and bath, referred to as the boundary mutual information (BMI). The corresponding entanglement entropy is captured via quantum extremal surfaces (QES), which holographically be computed by a surface optimization algorithm based on ``Surface Evolver''. We focus on both connected and disconnected configurations of the quantum entanglement wedge (Q-EW) in the AdS3_33​ bulk and analyze the finite corrections to the BMI. Our numerical results reveal a phase transition of the BMI as the separation between two subregions increases. Furthermore, we find that the BMI can naturally be decomposed into two distinct components: a geometric term arising from the areas of the quantum extremal surfaces, and a correction term resulting from bulk quantum fields within the Q-EW. Interestingly, the geometric contribution always exceeds the total BMI, indicating a negative correction from the bulk matter fields. This negativity can be understood as the result of subtracting a greater contribution from quantum fields in the connected Q-EW than in the disconnected one. We also reproduce the negative contribution of bulk quantum fields to BMI within a random tensor network (RTN) toy model of double holography. Modeling the bulk as a highly mixed state entangled with a large bath leads to a volume-law bulk entropy. In the large bond-dimension limit, the geometric part of the BMI remains non-negative, while the bulk entropy contribution becomes non-positive when the Q-EWs merge.