Digital Compute-in-Memory (DCiM) reduces data movement and has become a promising solution for energy-efficient edge AI. However, most existing DCiM frameworks still primarily target integer or fixed-point arithmetic, and provide limited support for compiler-integrated and accuracy-configurable floating-point computation. Directly integrating conventional IEEE 754 floating-point units into dense SRAM-based DCiM arrays, however, incurs high area and power overhead. To address this challenge, this work presents an accuracy-configurable floating-point multiplier integrated into the OpenACM framework for SRAM-based DCiM. An exact IEEE~754-compliant multiplier is first implemented as a baseline, and a mantissa-segmentation-based approximate multiplier is then proposed to reduce hardware cost while preserving numerical fidelity. Post-layout results show up to 69% logic area reduction and 72% power savings over exact floating-point designs without delay overhead. Evaluations on image processing tasks and ResNet-18 inference further demonstrate negligible accuracy degradation. These results indicate that compiler-integrated approximate floating-point multiplication is a practical approach for enabling efficient and configurable floating-point support in SRAM-based DCiM systems. The Floating-Point Multiplier is available on https://github.com/ShenShan123/OpenACM

Let $s$ denote West's stack-sorting map. In 2020, Defant characterized and enumerated the set $s^{n-m}(S_n)$ for $n \geq 2m-3$. While $|s^{n-m}(S_n)| = B_m$ when $n \geq 2m-2$, where $B_m$ denotes the $m$th Bell number, there are additional permutations when $n = 2m-3$. In this paper, we explore the more complex $n = 2m-4$ case, with several forms of additional permutations. We characterize $s^{m-4}(S_{2m-4})$ and find that its size is \[B_m + \frac{m^2 + 7m - 28}{2}\] for $m \geq 5$. This answers Defant's question about the $2m-4$ case. Furthermore, we find some differences in the behavior of the $2m-5$ case compared to the $2m-3$ and $2m-4$ cases.

Sustainable resource management requires harvesting strategies that account for environmental variability and ecological uncertainty. This study investigates optimal harvesting of renewable biological resources within a stochastic framework, where population dynamics are influenced by random environmental fluctuations and modeled using stochastic differential equations. Two complementary approaches are employed: the Hamilton-Jacobi-Bellman (HJB) equation and the Feynman-Kac representation. The HJB framework provides a dynamic optimization rule and characterizes the value function through a nonlinear partial differential equation, while the Feynman-Kac approach offers a probabilistic interpretation of expected returns. A comparative analysis demonstrates the theoretical consistency and practical relevance of both methods for designing economically efficient and ecologically sustainable harvesting policies under uncertainty.

It is a long-standing problem to prove that the number of distinct topological types of Calabi-Yau threefolds is finite. A related proposition, Reid's fantasy, conjectures that all Calabi-Yau threefolds are connected in a single moduli space through extremal transitions. Finiteness of topological types has been proven for the class of elliptic and genus one fibered Calabi-Yau threefolds, which recently have been shown to constitute the vast majority of known Calabi-Yau threefolds; the moduli space of elliptic CY3's is connected. In this letter, we demonstrate that all non-fibered Calabi-Yau threefolds in two of the largest known classes (toric hypersurfaces and complete intersections in products of projective spaces) are connected to fibered Calabi-Yau threefolds through a simple class of geometric transitions involving the shrinking of a single divisor from a fibered geometry. This suggests that non-fibered Calabi-Yau threefolds are rare special cases that are reached by simplifying fibered Calabi-Yau threefolds, and points to a natural path towards proving finiteness and Reid's fantasy for Calabi-Yau threefolds.

We present CritLens, a visual analytics system that helps users build personalized multi-criteria decision models from review text. In everyday decisions -- choosing equipment, hotels, or restaurants -- evaluation criteria are either preset by platforms or generated by LLMs, leaving users unable to discover, adjust, or verify them against the underlying evidence. This is problematic because many preferences are latent: they surface only upon encountering specific reviews, and any fixed framework risks overlooking low-frequency but decisive details. CritLens addresses this gap by using LLMs to transform reviews into an initial AHP decision model, then supporting iterative, human-in-the-loop refinement. Through coverage gap detection in the embedding space, users discover criteria missed by the initial model; through interactive weight adjustment under AHP consistency constraints, they express personal priorities; and through a multi-level scorecard and exportable decision report, they trace every ranking back to the original review text. Two case studies, an eight-participant user study, and a quantitative consistency-repair experiment demonstrate the system's effectiveness.

We consider nonlinear semidefinite programming problems with complementarity constraints (SDCMPCC), a class of highly degenerate problems where classical optimality conditions typically fail. In this context, weak stationarity conditions have been developed to address their degeneracy. While these notions are well understood, their algorithmic implications remain largely unexplored in semidefinite complementarity settings. We introduce a reformulation based on the spectral decomposition of the complementarity structure, which preserves local solutions and enables a tractable analysis. Within this framework, we analyze an augmented Lagrangian method for SDCMPCC and prove that under a suitable extension of Robinson's condition, every accumulation point of the generated sequence is W-stationary, or C-stationary under a stricter condition for solving the subproblems.

Two-dimensional superconductors with multiband electronic structures provide an ideal platform for exploring anisotropic and multigap superconductivity in the reduced-dimensionality limit. Here, we investigate the structural, electronic, vibrational, and superconducting properties of a hexagonal BAs$_3$ monolayer using first-principles calculations combined with density functional perturbation theory and fully anisotropic Migdal--Eliashberg theory. The optimized structure is found to be dynamically and thermally stable, as confirmed by phonon calculations and ab initio molecular dynamics simulations. Electronic structure calculations reveal an intrinsic metallic state with multiple bands crossing the Fermi level and several disconnected Fermi-surface sheets derived primarily from hybridized B-$p$ and As-$p$ orbitals. The electron--phonon interaction is dominated by low-frequency As-derived phonon modes, yielding a total electron--phonon coupling constant of $λ=0.75$. Solving the anisotropic Eliashberg equations predicts a superconducting critical temperature of $T_c=3.4$ K. The momentum-resolved superconducting gap exhibits a pronounced two-gap character with gap magnitudes of $Δ_1=0.75$ meV and $Δ_2=0.51$ meV at $T=1$ K. The superconducting gaps remain finite over the entire Fermi surface, demonstrating a fully gapped nodeless superconducting state. Analysis of the momentum-dependent electron--phonon coupling reveals that the two-gap superconductivity originates from sheet-dependent pairing interactions associated with distinct Fermi-surface sheets. Our results establish monolayer BAs$_3$ as an intrinsic anisotropic two-gap superconductor and expand the growing family of boron-based two-dimensional superconductors.

We characterize the equality case in the Brunn-Minkowski inequality for the Monge-Ampère eigenvalue of general bounded convex domains. This together with the author's previous work completely resolves a question raised by Salani (A Brunn-Minkowski inequality for the Monge-Ampère eigenvalue. Adv. Math. 194 (2005)).

We study common-value auctions in which bidders have horizontally differentiated preferences. In a specific two-bidder parameterization, winning conveys good news about the object's value to the winner, a phenomenon we call the winner's bliss in contrast to the conventional winner's curse. Additional implications also differ from the conventional analysis. When bidders' preferences are horizontally differentiated, seller revenue is reduced with information disclosure, and advantageous selection sustains bilateral trade under asymmetric information.

Many model evaluation tasks reduce to estimating an average loss, error rate, or subgroup metric on a stratified pool when each label, human rating, or simulator call is costly. The precision-optimal Neyman allocation depends on within-stratum variances, which must be learned from the same observations used for estimation. We formulate this as a sequential allocation problem and use the exact one-step marginal variance reduction as the priority index. Replacing the unknown variances by independent inverse-chi-squared posterior draws yields TS-Neyman, a Thompson-sampling rule that preserves the oracle marginal-gain structure while randomizing over variance uncertainty. For any fixed finite number of strata, we prove almost-sure convergence of the TS-Neyman allocation proportions to the Neyman target, asymptotic optimality of the variance proxy, and a central limit theorem for the resulting adaptive stratified estimator. In two five-stratum budget-scaling benchmarks, one bounded-loss benchmark and one binary model-error benchmark in the spirit of Dai et al. 2023, TS-Neyman's relative efficiency stays within 5 percent of the oracle on the bounded-loss population and within about 15 percent on the binary benchmark. In an additional CivilComments real-data replay with confidence-based strata, it stays within about 8 percent of the oracle and improves on equal allocation by roughly 7 to 14 percent in MSE across budgets, while plug-in greedy and two-stage plug-in can degrade by over an order of magnitude under sparse pilots. Common-pilot warm-start and prior-sensitivity studies show that this behavior is stable under working-model and working-prior misspecification.

The $abc$-conjecture (Masser and Oesterle) has remained open for decades. By measuring $abc$-triples using a particular quality metric, the conjecture may be framed as seeking the asymptotic distribution of triples of sufficient quality. We create new classes of quality metrics to develop variants on the $abc$-conjecture, with each metric based upon the doubly geometric mean of the prime factors of triples. We investigate the behavior of the resulting class of quality metrics; by determining families of triples that yield high quality, we establish several asymptotic results that are analogous to the $abc$-conjecture for our metrics. We also develop sharp phase transitions for the behavior of families of such quality metrics within specified parametrizations for smoothness of primes in $abc$-triples, using heuristics from the Szpiro ratio for associated Frey curves. Finally, we implement algorithms to determine triples with high qualities with sub-linear runtime, an asymptotic speedup over naïve approaches. Our analysis offers robust variations of, and connections to, the $abc$-conjecture that offer independent questions of analytical interest.

We present a structured analysis of how contemporary clinical AI systems integrate electronic health record (EHR) data and the extent to which they support longitudinal clinical reasoning. Drawing on a curated corpus of clinical natural language processing (NLP) and EHR-integrated systems, we develop a coding framework that captures both technical integration strategies and reasoning-relevant representational features, such as trajectory modeling, cross-encounter synthesis, longitudinal analysis, and absence reasoning. We also elicited the experiences of three physicians in their EHR use, including what strengths and weaknesses they found with their institution's current EHR system(s). Our analysis shows that while many systems incorporate EHR data, they predominantly operate on encounter-level or aggregated representations, with limited support for explicit temporal reasoning across patient histories. Reasoning-relevant structures are inconsistently represented, and evaluation paradigms remain largely focused on predictive performance instead of longitudinal interpretability. We argue that current approaches treat EHR data as a static input rather than a substrate for ongoing clinical reasoning, and we outline a framework for understanding how future systems might more effectively align with the temporal and interpretive structure of clinical practice.

Given a finite set of integers $A$, a \emph{unary translocation} produces a new set $A' = A \cup \{u,v\}$, where $u$ and $v$ are nonnegative integers satisfying $x+y=u+v$ for some $x,y\in A$. For an input set $A$ and a target set $B$, the \emph{unary translocation distance} is the minimum number of unary translocations required to obtain a superset containing $B$. In this paper, we study this problem from both theoretical and computational perspectives. We prove that computing the unary translocation distance is strongly NP-hard, thereby answering an open question raised by \citet{ConstantinMiclausPopa2026UnaryTranslocation}. On the positive side, we give an exact pseudo-polynomial algorithm for every fixed constant value of $|B|$, extending our previous results for $|B|\leq 2$. For arbitrary target sets, we present a $2$-approximation algorithm, an additive $(|B|-1)$-approximation algorithm, and show that the additive algorithm also yields a $3$-approximation. We also propose parameterized algorithms, including algorithms parameterized by the maximum value in the input set together with the optimum distance, and by the maximum value in the target set together with $|B|$. In addition, we propose an integer linear programming formulation that gives an exact mathematical model for the problem, analyze its size, and show that the LP relaxation has integrality gap at least $\frac{4}{3}$. Finally, we report computational experiments comparing the $2$-approximation algorithm, beam search, and simulated annealing. The results show that the approximation algorithm is highly effective in practice and often outperforms the heuristic baselines.

Phylogenetic networks are graphs inferred from molecular sequence data that represent ancestral histories shaped by reticulate processes such as recombination, hybridization, and horizontal gene transfer. We introduce a family of distance metrics for rooted, ranked, unlabeled phylogenetic networks, extending a previously developed distance for ranked trees. Our approach relies on a bijective triangular matrix representation of phylogenetic networks that captures the temporal order of internal events, speciations, and hybridizations. Our metrics, defined as standard matrix norms, allow efficient quantitative comparisons of network topologies, timed networks and networks with differing numbers of hybridizations. Our distance can be used for both isochronous networks where all tips are sampled at one time point, and heterochronous networks where tips are allowed to be sampled at different time points. We show that our metrics capture biologically meaningful differences among evolutionary histories in both simulations and empirical posterior distributions of viral phylogenetic networks. These tools fill a methodological gap, enabling principled comparisons of ranked, unlabeled phylogenetic networks, including ancestral recombination graphs.

Characterizing directed information flow in brain networks is difficult because neural circuits are full of recurrent feedback loops. Many existing tools for directed dependence assume a directed acyclic graph (DAG) structure to resolve directional ambiguity, and therefore cannot represent these loops. We present a nonparametric, information-theoretic framework that addresses this by coupling the discrete Hodge decomposition with lead-lag mutual information, splitting the resulting edge flow into three orthogonal components: a gradient term capturing hierarchical, feed-forward relationships; a curl term isolating triangle-level feedback loops; and a harmonic term capturing cyclic flow around topological holes. This separation makes it possible to disentangle feed-forward drive from recurrent circulation, which conventional measures conflate. We further develop a permutation-based hypothesis-testing layer that identifies nodes and triangular motifs whose information-flow signatures change significantly between conditions. We validate the framework on simulations with known ground-truth structure and apply it to local field potential recordings from a rodent model of focal ischemic stroke. In three of four animals, we find a post-stroke shift toward hierarchical, source-driven propagation at the expense of recurrent feedback, while the fourth shows no significant change.

Advancing superconducting quantum devices to higher operating frequencies broadens their functionality and enables operation at elevated temperatures, but it also requires near-quantum-limited amplifiers beyond the few-gigahertz regime. Here we present a junction-free, kinetic-inductance parametric amplifier based on thin-film niobium nitride (NbN) operating at 23 GHz in the microwave K-band, achieving a gain up to 40 dB, a 100 MHz gain-bandwidth product, a 1 dB saturation input power of -85 dBm with 23 dB gain, and added noise no greater than 1.4 quanta for phase-preserving amplification. Leveraging the large superconducting gap of NbN, this architecture can be extended to even higher frequencies, supporting applications such as high-fidelity readout of millimeter-wave superconducting qubits and axion searches over an expanded mass window.

For a digraph $D$, let $\aTwo(D)$ be the largest size of a vertex set no two of whose vertices lie in a common directed $2$-cycle. Let $f_2(a)$ be the least integer $K$ such that every $K$-connected digraph $D$ with $\aTwo(D)\leq a$ has a Hamilton cycle. In 1987, Jackson proved that $f_2(a)\leq 2^a(a+2)!$ and asked for better bounds, noting that a linear bound might be possible. Kühn and Osthus later observed that even a polynomial bound would be interesting. In this short note, we prove the polynomial bound $f_2(a)\leq 2a^3+2$.

HgTe quantum dots are promising mid-infrared nanomaterials owing to their exceptional bandgap tunability, yet their electronic structure is strongly influenced by surface coordination and ligand passivation at ultrasmall sizes. Here, we employ atomistic simulations to systematically investigate stoichiometric HgTe nanoclusters with sizes 0.86 to 1.85 nm. The in silico exploration uncovers a transition from confinement-dominated electronic structures with delocalized frontier states in small self-passivated clusters to surface influenced characteristics in larger nanoclusters. Increased coordination and bond-length inhomogeneity in the larger nanoclusters generate localized near-gap states centered on undercoordinated surface atoms. At intermediate sizes, the band edge states become spatially separated on different regions of the cluster without forming deep gap states, marking the onset of surface induced electronic asymmetry. In larger clusters (1.8 nm), common neutral ligands like amines, thiols, phosphines, and alcohols effectively eliminate surface-derived localized states by restoring local coordination and altering the band edge electronic structure through ligand surface hybridization. The sensitivity of the bandgap to ligand identity and binding site underscores the interplay between surface coordination and ligand chemistry in shaping the electronic structure of these nanoclusters. These insights provide an atomistic understanding of size-dependent electronic structures in ultrasmall HgTe clusters. The study further establishes neutral ligands as powerful chemical handles for engineering frontier electronic states relevant to infrared optoelectronic functionality.

We show analytically and via simulation that cross-sectional aggregation can substantially attenuate regime-switching signals in time-series data, making regime switches harder to detect. Building on this, we develop regime-switching models and an estimation algorithm which allow for autoregressive dynamics and grouped heterogeneity. We apply the approach to a U.S. macroeconomic dataset of 94 series, covering components of real gross domestic product, industrial production, capacity utilization, employment, and hours worked. The estimates give sharper business cycle classifications than those typically found in the literature. Monte Carlo simulations show that the computation is practical for datasets with a few hundred time series.

We develop an explicit PBW normal ordering theory for the $26$ double extension regular algebras of type $(14641)$ in the Zhang-Zhang classification. With respect to the order $x_1\prec x_2\prec y_1\prec y_2$, we obtain closed two-letter formulas for the internal relations and recursive coefficient systems for mixed words, products of PBW monomials, powers of normal blocks, and noncommutative multinomial expressions. The internal coefficients are mostly quantum or skew-commutative, while the Jordan families produce Lah-Whitney, hence Stirling-type, triangular arrays. The symbolic reductions are supported by a SageMath implementation included as an ancillary file.

Video-to-audio (V2A) generation must jointly satisfy audiovisual alignment, semantic consistency, temporal synchronization, and perceptual quality. While prior work has mainly focused on model architecture, multimodal conditioning, and training objectives, inference-time alignment for V2A remains underexplored. In this paper, we study inference-time alignment for flow-matching-based V2A generation and formulate it as a search problem. We propose Sequential Monte Carlo Inference-Time Alignment (SMC-ITA), which combines lookahead-based reward estimation and sequential Monte Carlo resampling to reallocate computation adaptively using multi-dimensional cross-modal rewards. SMC-ITA improves over naive single-trajectory sampling, achieving a 55.67% relative reduction in DeSync, a 20.23% improvement in IB-score, and a 15.44% improvement in Audio Quality. Under matched NFE budgets, it also achieves the best overall trade-off among the compared search baselines, outperforming Best-of-N and Beam Search. Ablation studies further show that lookahead improves the reliability of intermediate reward estimates and that systematic resampling is a strong practical default for V2A inference-time alignment.

We introduce the notion of an ``initial condition'' for a module over a commutative Noetherian local ring, allowing for a recursive construction of its ``solution modules''. If the given module has zero-dimensional support, such as the residue field of the local ring and those encountered in residual complexes, we demonstrate that the solution module is an injective hull of the given module. The construction of the solution module for finitely generated given module is explicit and computable, devoid of the need for Zorn's lemma.

In April 2026, the MV Hondius expedition cruise ship became the site of the first documented cruise ship-associated Andes hantavirus (ANDV) cluster, with 13 confirmed and probable cases and 3 deaths among 149 passengers and crew. We applied a stochastic epidemic model to evaluate four embarkation scenarios under reproductive numbers anchored to published ANDV estimates. Scenario D, involving two latent infected persons at embarkation, was most consistent with the observed outbreak, yielding P(final size >= 13) = 11.6% and P(takeoff) = 58.5% at R0 = 2.12. Approximate Bayesian computation provided complementary support for multiple latent infections at embarkation, especially E1(0)=1 and E3(0)=2, but R0 remained weakly identifiable. A day-35 transmission reduction changed takeoff probability little in this counterfactual model. Findings support exposure-history assessment, early onboard surveillance, rapid isolation of symptomatic cases, and postdisembarkation monitoring for travelers from ANDV-endemic regions.

In this paper, we mainly focus on pointwise, arbitrarily high-order interior estimates for the mixed local-nonlocal elliptic equation \begin{equation*} (-Δ)^{s}u(x)-Δu(x)=f(x),\quad x\in B_r(0) \end{equation*} with $0<s<1$. The main challenges in this setting are the absence of an explicit Green function and the ineffectiveness of standard bootstrap arguments. Our approach overcomes these difficulties via the Campanato iteration method, which inductively constructs polynomials approximating the solution. Using the fact that all functions are locally $s$-harmonic up to a small error significantly reduces the computational complexity when studying the regularity of the fractional Laplacian acting on these polynomials. Finally, the regularity estimates are obtained from the analysis of a manageable recursive inequality system.

In this note we describe a systematic procedure for computing partition functions of noninvertible Rep ($G$) quantum symmetry actions on two-dimensional nonabelian $G$ orbifolds. We apply this procedure to multiplicity-free examples for several nonabelian groups, and check explicitly in those examples that the partition function of a two-dimensional Rep ($G$)-gauged $G$ orbifold, for nonabelian $G$, matches that of the original theory, as expected on general principles. This fills a minor gap in the literature, making quantum symmetries in two-dimensional nonabelian orbifolds more concrete.

In this paper, we study variational properties of the metric projection mapping onto isotone projection cones in finite-dimensional Euclidean spaces. We derive explicit formulas for both the Fréchet coderivative and the Mordukhovich coderivative of the projection operator. The analysis is based on a local description of the projection mapping via an associated generating system in a neighborhood of a reference point, which leads to computable coderivative characterizations. As an application, we compute the covering constant of the projection mapping, providing a quantitative description of its local regularity. Furthermore, we establish verifiable sufficient conditions for the Aubin property of the solution mapping associated with parametric nonlinear complementarity problems associated with isotone projection cones. The obtained results contribute to the variational analysis of metric projections in a general cone setting where orthogonality arguments are not available, and to the stability theory of complementarity systems in finite dimensions.

We produce a model of ZF + DC in which there exists a discontinuous homomorphism from the real line to itself but no Hamel basis for the real line, and prove a generalization of this result in terms of internal direct sums.

Recent DESI BAO observations provide intriguing hints that dark energy may be dynamical in nature. To investigate deviations of the dark energy equation of state (EoS) from $w = -1$, we introduce the $δ$-CDM framework, a controlled deformation of $Λ$CDM in which deviations from a cosmological constant are parametrized by a redshift-dependent function $δ(z)$, defined through $w_{\rm de}(z) = -1 + δ(z)$. As an illustrative example, we reconstruct $δ(z)$ using effective scalar field dynamics of thawing type, encompassing both quintessence and phantom regimes within a unified description. Notably, the reconstructed $δ(z)$ is independent of the specific scalar field realization, ensuring theoretical robustness. Using Planck CMB-SPA data, DESI DR2 BAO measurements, and the Pantheon+ supernova sample within a Bayesian Markov Chain Monte Carlo analysis, we find that the $\tilde{w}_0\tilde{w}_a$ parametrization is preferred over this thawing-type realization of deviations from $w = -1$. Overall, the $δ$-CDM framework provides a minimal yet flexible extension of $Λ$CDM, capable of capturing late-time dynamical features of dark energy.

Domain-specific Field Programmable Gate Array (FPGA) architectures increasingly integrate specialized hardblocks, such as Tensor Slices, to accelerate artificial intelligence and machine learning workloads. Despite their efficiency benefits, these architectures remain difficult to program because designers typically rely on manual Register-Transfer Level (RTL) integration to access these hardblocks. This paper presents a compiler-agnostic methodology that enables high-level synthesis (HLS) tools to target custom FPGA hardblocks directly from C/C++ code. Architectural hardblocks are exposed as schedulable C-level operators using an RTL blackbox abstraction with explicit latency and initiation-interval contracts, allowing the HLS scheduler to optimize around specialized hardware without manual RTL orchestration. Unlike traditional uses of HLS blackboxes for external IP integration, our approach treats blackboxes as architectural abstractions, enabling scalable composition of C-level operators that target custom FPGA hardblocks without compiler modification. We evaluate the proposed flow using a Tensor Slice-based FPGA architecture with AMD Vitis HLS and the Verilog-to-Routing (VTR) toolchain. Across multiple matrix sizes, designs generated using the proposed C-Blackbox flow achieve lower area-delay product than behavioral HLS baselines while providing substantially higher productivity-adjusted efficiency than handwritten RTL implementations. These results demonstrate that domain-specific FPGA architectures can be made accessible through HLS while maintaining competitive hardware efficiency.

In this work, we report a systematic study of the impact of high-temperature post-capping thermal annealing on the memory characteristics of Oxide-semiconductor channel ferroelectric field-effect transistors (OS-FeFETs). Using an identical engineered ferroelectric gate stack 8nm Hf0.5Zr0.5O2 (HZO) / 3 nm Al2O3 / 8 nm HZO (8/3/8) and a hybrid capping layer (3 nm HfO2 + 3 nm Al2O3), 10 percent Ga doped InO (IGO) channel and 4 percent W doped InO (IWO) channel FeFETs remain functional after annealing at temperatures up to 650 C for durations of up to 30 min and 10 min, respectively; further annealing results in irreversible loss of conduction and device failure. Detailed electrical analysis reveals that the MW enhancement originates from a preferential positive shift in the erased-state threshold voltage, while the programmed-state threshold voltage remains comparatively stable. Grazing-incidence X-ray diffraction measurements further indicate structural evolution in the IWO and IGO oxide channels with increasing annealing temperature, supporting the observed electrical trends.