Mixed-Primitive-Based Initialization

• Mixed-primitive-based initialization is a design pattern that composes multiple primitive components to create structured, flexible starting states for various optimization problems.
• In parametrized quantum circuits, the approach avoids barren plateaus by leveraging tailored gatewise strategies and biasing parameter distributions to improve gradient and convergence properties.
• Across domains like motion planning, 3D reconstruction, and placement, the strategy balances computational efficiency and accuracy by matching initialization biases to the specific geometric and operational constraints.

Searching arXiv for the cited topic and related initialization papers. to=arxiv_search.search 新天天彩票={"^^^^^^^^6query6 Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^^^^^", "^^^^^^^^6max_results6 ^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^} code omitted? Searching for the main barren-plateau paper and related initialization work. to=arxiv_search.search 尚度={"^^^^^^^^6query6 OR (Grant et al., 2019) OR \6^^^^^^^^"Exponentially many initializations to avoid barren plateaus\"^^^^^^^^6 OR \6^^^^^^^^"An initialization strategy for addressing barren plateaus in parametrized quantum circuits\"", "^^^^^^^^6max_results6 ^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^} code omitted? to=arxiv_search.search аанацҳауеит 北京赛车开奖իչ 天天中彩票怎么买 code չկա? to=arxiv_search.search 天天彩票怎么 彩神争霸是 code to=arxiv_search.search ￣奇米影视ीक्षक code to=arxiv_search.search 代理娱乐 code to=arxiv_search.search code 天天中彩票追号json {"^^^^^^^^6query6 \"Exponentially many initializations to avoid barren plateaus\" (&&&^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^&&&)", "^^^^^^^^6max_results6 ^^^^6 OR \6^^^^} to=search_arxiv code Mixed-primitive-based initialization denotes a family of initialization procedures in which the starting state of an optimization or generative system is assembled from multiple primitive components rather than from a single homogeneous prior. In recent arXiv literature, the relevant primitives are domain-specific: gatewise parameter distributions in parametrized quantum circuits, trajectory prototypes in truncated diffusion planners, simplex-shaped splatting elements in surface reconstruction, superquadric assemblies for ^^^^6query6^^^^D Gaussian Splatting, and discrete reasoning instructions for LLM warm starts. Across these settings, the common objective is to bias the initial state toward trainable, feasible, or diverse regions of the landscape while retaining sufficient flexibility for subsequent optimization (&&&^^^^6query6^^^^&&&, Xu et al., 5 Apr 2026, Qu et al., 15 Jul 2025, Huc et al., 25 Jul 2025, &&&^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^&&&).

^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^. Cross-domain definition and taxonomy

A mixed-primitive initializer is characterized by two ingredients: a library of primitives and a rule for composing them into an initial state. The library may be discrete and small, as in the two-Gaussian gatewise constructions used for barren-plateau avoidance, or structured and learned, as in primitive prediction heads for motion planning and ^^^^6query6^^^^DGS. The composition rule may be independent sampling, deterministic inverse pairing, clustering-based assignment, top-PRESERVED_PLACEHOLDER_^^^^6query6^^^^ selection, or weighted mixing.

Domain	Primitive set	Initialization mechanism
Variational quantum circuits	PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^, PRESERVED_PLACEHOLDER_^^^^6max_results6^^^^, shifted or windowed parameter laws	Gatewise selection; identity-block inverse pairing
Differential-drive motion planning	PRESERVED_PLACEHOLDER_^^^^6query6^^^^ trajectory primitives	Truncated diffusion anchored on a selected primitive
Gaussian Splatting surface reconstruction	Gaussian ellipse, Gaussian line, Gaussian triangle	Cluster-size assignment PRESERVED_PLACEHOLDER_^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^-simplex, PRESERVED_PLACEHOLDER_^^^^6 OR \6^^^^-simplex, $3\!\to\!2$-simplex
Spacecraft ^^^^6query6^^^^DGS	$M$ superquadric primitives	CNN predicts primitive assembly and pose
LLM warm start for RL	$Z=\{z_1,\dots,z_m\}$ reasoning primitives	Primitive sampling with weights $\alpha_z$

A related formulation appears in mixed-size global placement, where initialization is not expressed as a primitive library in the geometric sense, but as a co-optimization between area-hint refinement and macro-schedule restoration. There, virtual macro-nodes, bin-nodes, and signed edges act as structured ingredients that bridge point-based and area-aware initialization regimes (&&&^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^&&&).

This suggests a unifying interpretation: mixed-primitive initialization is less a single algorithm than a design pattern for replacing isotropic, unstructured starts with structured priors that preserve multiple admissible modes.

^^^^6max_results6^^^^. Operator-level formulations in parametrized quantum circuits

In parametrized quantum circuits, mixed-primitive initialization has been developed as a response to barren plateaus. A first-moment framework introduces an operator-level diagnostic for whether an initialization can escape the fully concentrated fixed point. If PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^ denotes a naïve barren-plateau initialization and PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^ an alternative, then a necessary criterion is

PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6^^^^

For a single gate PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^ with PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^, the first moment is governed by

PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6 OR \6^^^^

and the mean gradient satisfies

PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)66^^^^

This immediately distinguishes symmetric and biased primitives. Identity initialization has PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^7, zero-centered Gaussian PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^8 has PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^9, shifted Gaussian PRESERVED_PLACEHOLDER_^^^^6max_results6query6^^^^ has PRESERVED_PLACEHOLDER_^^^^6max_results6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^ and PRESERVED_PLACEHOLDER_^^^^6max_results6max_results6^^^^, and a uniform window PRESERVED_PLACEHOLDER_^^^^6max_results6query6^^^^ yields PRESERVED_PLACEHOLDER_^^^^6max_results6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^, PRESERVED_PLACEHOLDER_^^^^6max_results6 OR \6^^^^, PRESERVED_PLACEHOLDER_^^^^6max_results66^^^^. The same framework shows that many shifted, biased, and non-symmetric parameter distributions can avoid concentration, and that one can construct PRESERVED_PLACEHOLDER_^^^^6max_results6^^^^7 inequivalent gatewise strategies by independently choosing between a zero-centered Gaussian and a PRESERVED_PLACEHOLDER_^^^^6max_results6^^^^8-centered Gaussian at each of PRESERVED_PLACEHOLDER_^^^^6max_results6^^^^9 gates. With variance PRESERVED_PLACEHOLDER_^^^^6query6query6^^^^, PRESERVED_PLACEHOLDER_^^^^6query6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^, so each choice escapes concentration. In the MaxCut example, the gap between two binary choices PRESERVED_PLACEHOLDER_^^^^6query6max_results6^^^^ satisfies PRESERVED_PLACEHOLDER_^^^^6query6query6^^^^, and the numerics in Fig. 8 show distinct final minima and bimodal histograms for different PRESERVED_PLACEHOLDER_^^^^6query6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^-families (&&&^^^^6query6^^^^&&&).

An earlier and more specialized quantum construction is the identity-block initializer. There, a circuit of total depth PRESERVED_PLACEHOLDER_^^^^6query6 OR \6^^^^ is partitioned into PRESERVED_PLACEHOLDER_^^^^6query66^^^^ contiguous blocks, each containing PRESERVED_PLACEHOLDER_^^^^6query6^^^^7 primitive gates,

PRESERVED_PLACEHOLDER_^^^^6query6^^^^8

with PRESERVED_PLACEHOLDER_^^^^6query6^^^^9 chosen so that PRESERVED_PLACEHOLDER_^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6query6^^^^. Each block evaluates exactly to the identity, so PRESERVED_PLACEHOLDER_^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^. The key theoretical point is that varying a single parameter only exposes an effective depth of at most PRESERVED_PLACEHOLDER_^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6max_results6^^^^, preventing the derivative circuit from behaving as a ^^^^6max_results6^^^^-design when PRESERVED_PLACEHOLDER_^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6query6^^^^. The paper states that random initialization gives PRESERVED_PLACEHOLDER_^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^, whereas identity-block initialization gives PRESERVED_PLACEHOLDER_^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6 OR \6^^^^, independent of PRESERVED_PLACEHOLDER_^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \66^^^^. Its reported initialization variances for PRESERVED_PLACEHOLDER_^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^7 remain near PRESERVED_PLACEHOLDER_^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^8 from PRESERVED_PLACEHOLDER_^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^9 to PRESERVED_PLACEHOLDER_^^^^6 OR \6query6^^^^, while random initialization drops from PRESERVED_PLACEHOLDER_^^^^6 OR \6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^ to PRESERVED_PLACEHOLDER_^^^^6 OR \6max_results6^^^^. On a PRESERVED_PLACEHOLDER_^^^^6 OR \6query6^^^^-qubit MNIST QNN with PRESERVED_PLACEHOLDER_^^^^6 OR \6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^, training accuracy reaches PRESERVED_PLACEHOLDER_^^^^6 OR \6 OR \6^^^^ in PRESERVED_PLACEHOLDER_^^^^6 OR \66^^^^ iterations with identity-block initialization versus PRESERVED_PLACEHOLDER_^^^^6 OR \6^^^^7 with zero initialization; on a PRESERVED_PLACEHOLDER_^^^^6 OR \6^^^^8-qubit Heisenberg VQE with PRESERVED_PLACEHOLDER_^^^^6 OR \6^^^^9 plus fixed entangling $3\!\to\!2$^^^^6query6^^^^, the final energy error is $3\!\to\!2$^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^ with identity initialization and no convergence with zero or random initialization (&&&^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^&&&).

Taken together, these results reject the simplified claim that there is a single “careful initialization” that cures barren plateaus. The mixed-primitive perspective instead emphasizes a large equivalence class of non-concentrating starts, each with its own first-moment bias.

^^^^6query6^^^^. Truncated diffusion anchored on trajectory primitives

In motion planning for differential-drive mobile manipulators, mixed-primitive initialization appears as primitive-based truncated diffusion. The primitive library is built offline from dynamically feasible trajectories

$3\!\to\!2$^^^^6max_results6^^^^

followed by $3\!\to\!2$^^^^6query6^^^^-Means in fixed-length path space with $3\!\to\!2$^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^, yielding $3\!\to\!2$^^^^6 OR \6^^^^ centroids $3\!\to\!2$6. Boundary states are encoded separately as ordered ^^^^6query6^^^^D keypoint sequences extracted via differentiable forward kinematics, and point clouds and keypoints are fused with attention in the task representation encoder (Xu et al., 5 Apr 2026).

The initialization itself replaces pure-Gaussian diffusion starts with a truncated and anchored prior. Instead of beginning from noise at $3\!\to\!2$7, the method truncates at $3\!\to\!2$8 and samples

$3\!\to\!2$9

where $M$^^^^6query6^^^^ is drawn from the primitive library. A primitive head predicts logits $M$^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^, followed by a softmax over the $M$^^^^6max_results6^^^^ primitives. At inference, one either samples $M$^^^^6query6^^^^ primitive indices from $M$^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^ or takes the top-$M$^^^^6 OR \6^^^^ indices; for each selected primitive, one samples $M$6, forms

$M$7

and runs $M$8 reverse DDIM or DDPM steps. The full planner then prunes collisions, initializes the piecewise-polynomial optimizer, optimizes candidates in parallel, and returns the best collision-free, dynamically feasible trajectory (Xu et al., 5 Apr 2026).

The reported motivation is threefold. Efficiency arises because a plausible primitive allows very few reverse steps, with the paper giving “e.g. ^^^^6max_results6^^^^” steps versus a $M$9–$Z=\{z_1,\dots,z_m\}$^^^^6query6^^^^-step vanilla DDPM. Diversity comes from sampling different primitives, which is said to preserve different homotopy classes. Optimality comes from using each primitive as a sensible prior path that denoising refines locally. Empirically, PTDM with $Z=\{z_1,\dots,z_m\}$^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^ samples achieves higher success rate, improved trajectory diversity, and competitive runtime relative to vanilla diffusion and classical baselines; it also outperforms both vanilla DDPM and the anchor-based diffusion of Liao et al. in mIoU-based diversity score and planning time (Xu et al., 5 Apr 2026).

^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^. Geometric initialization in Gaussian Splatting and ^^^^6query6^^^^DGS

In surface reconstruction, mixed-primitive initialization is used to address the representation gap created by a single splatting primitive. One framework introduces three geometric primitives: Gaussian ellipse as a $Z=\{z_1,\dots,z_m\}$^^^^6max_results6^^^^-simplex, Gaussian line as a $Z=\{z_1,\dots,z_m\}$^^^^6query6^^^^-simplex, and Gaussian triangle as a $Z=\{z_1,\dots,z_m\}$^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^-simplex. All carry an opacity $Z=\{z_1,\dots,z_m\}$^^^^6 OR \6^^^^, spherical-harmonic color coefficients $Z=\{z_1,\dots,z_m\}$6, and covariance $Z=\{z_1,\dots,z_m\}$7, but differ in vertex structure. Initialization proceeds from a COLMAP point cloud $Z=\{z_1,\dots,z_m\}$8 by partitioning points into clusters $Z=\{z_1,\dots,z_m\}$9 satisfying

$\alpha_z$^^^^6query6^^^^

The clustering tree is built by distance-based hierarchical clustering, then traversed by BFS to collect color-consistent subsets of size at most $\alpha_z$^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^. Primitive type is determined directly by cluster size: $\alpha_z$^^^^6max_results6^^^^ gives an ellipse, $\alpha_z$^^^^6query6^^^^ a line, and $\alpha_z$^^^^6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^ a triangle. For lines and triangles, local orthonormal frames are constructed from the clustered points, and local coordinates $\alpha_z$^^^^6 OR \6^^^^ are stored in that frame. The downstream optimization then interleaves compositional splatting, photometric loss $\alpha_z$6, depth losses $\alpha_z$7, regularization on $\alpha_z$8 and $\alpha_z$9, and primitive- and vertex-level pruning plus densification. The paper argues that matching local simplex dimension to local geometry reduces the “representation gap” of a one-size-fits-all ellipse model and yields higher fidelity surfaces with fewer artifacts (Qu et al., 15 Jul 2025).

A related but distinct mixed-primitive initializer appears in pose-aware ^^^^6query6^^^^D Gaussian Splatting for non-cooperative spacecraft. There, the primitive library is a union of PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6query6^^^^ superquadrics with parameter vectors

PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^

where PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6max_results6^^^^ are shape exponents, PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6query6^^^^ are half-axis lengths, PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^ is translation, PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6 OR \6^^^^ is the 6D rotation representation, and PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query66^^^^ are taper parameters. A ResNet-^^^^6query6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^ backbone outputs a PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^7-dimensional feature vector, followed by a shape head predicting PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^8 and a pose head predicting camera rotation and translation; the total learnable parameter count is PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^9. The CNN is trained with rotation, translation, reprojection, and Chamfer point-cloud losses. To initialize ^^^^6query6^^^^DGS, the method samples surface points from each predicted superquadric, maps them to world coordinates PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^, and creates one Gaussian per sample with covariance PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^, PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6^^^^, small constant opacity PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^, and color taken from the projected input pixel. The reported effect is substantial acceleration of downstream ^^^^6query6^^^^DGS. Using ground-truth poses, the number of iterations required to reach PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^ falls from PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6 OR \6^^^^ for random initialization to PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)66^^^^ for CNN Original, PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^7 for CNN Ambiguity-Aware, and PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^8 for CNN Ambiguity-Free; the number of images drops from PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^9 for random initialization to PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6query6^^^^ for the CNN-based initializations. Under CNN-estimated poses, only the ambiguity-free network is reported to have sufficiently low angular errors, PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^ on train and PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6max_results6^^^^ on test, for recovery of solar-panel geometry (Huc et al., 25 Jul 2025).

These reconstruction papers define “mixed primitive” geometrically rather than probabilistically. The initializer is not merely a better seed for the same parameterization; it changes the initial representational basis.

^^^^6 OR \6^^^^. Mixed primitives in placement and reasoning warm starts

In mixed-size global placement, initialization is framed as a co-optimization problem between a fast point-based start and a more faithful area-aware objective. The area-hint refinement initializer augments the netlist graph PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6query6^^^^ with two kinds of virtual nodes: one macro-node for each fixed macro and one bin-node for each bin in a PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^ partition of the placement region. Macro-repulsive and bin attraction/repulsion edges are added with signed weights. For a macro PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6 OR \6^^^^, the negative edge weight is

PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results66^^^^

while for a bin PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6^^^^7,

PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6^^^^8

These define a signed adjacency PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6^^^^9 and signed Laplacian PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6query6^^^^. Starting from a raw GSP initialization PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^, refinement applies the signed-graph filter

PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6max_results6^^^^

inside a short fixed-point iteration

PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6query6^^^^

The second stage, macro-schedule placement, progressively restores macro area constraints by evolving each macro’s charge distribution from a point-like model toward a full footprint using redistribution or restoration schedules. On PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^ ISPD^^^^6max_results6query6query6 OR \6^^^^ mixed-size circuits and two industrial designs under two nodes, for PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6 OR \6^^^^ cases in total, the method improves HPWL over point-based initializers in PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query66^^^^ out of PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^7 cases, achieves up to PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^8 HPWL reduction, runs approximately PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^9 times faster than the QCQP area-aware initializer, and keeps total global placement runtime within PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6query6^^^^–PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^ of DREAMPlace. Among four macro schedules, exponential-like restoration performs best on most circuits; robustness is reported across target densities PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6max_results6^^^^–PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6query6^^^^ and random seeds with PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^ variation PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6 OR \6^^^^ (&&&^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^&&&).

In reinforcement learning for LLM reasoning, mixed primitives take the form of discrete reasoning instructions. The underlying MDP is PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \66^^^^, and a primitive PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^7 is a short instruction or prompt suffix such as a top-down or bottom-up strategy. Under a primitive PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^8, the policy is

PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^9

The resulting thinking-state distribution is PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6 OR \6query6^^^^, and if primitives are mixed with weights PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6 OR \6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^, then

PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6 OR \6max_results6^^^^

The Tailor pipeline consists of demonstration trace analysis, reasoning primitive synthesis, Tailor SFT dataset curation, and SFT followed by RL. In SFT curation, the teacher is prompted to solve a seed ^^^^^^^^6query6^^^^^^^^ using a sampled primitive PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6 OR \6query6^^^^, and the student is trained on the resulting trajectories via standard SFT loss. Uniform mixing is the default, but PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6 OR \6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^ can also be set proportional to PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6 OR \6 OR \6^^^^ or PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6 OR \66^^^^. The RL stage uses KL-regularized PPO or DAPO. Empirically, the paper reports that on Knights-and-Knaves, reaching PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6 OR \6^^^^7 accuracy requires PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6 OR \6^^^^8 fewer RL steps than any baseline, and that PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6 OR \6^^^^9 is roughly PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)66query6^^^^–PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)66Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^ lower than the best baseline on KK and iGSM. Final accuracies are PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)66max_results6^^^^ in-distribution and PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)66query6^^^^ OOD on KK, PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)66(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^ in-distribution and PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)66 OR \6^^^^ OOD on iGSM-Medium, and PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)666^^^^ on iGSM-Hard (&&&^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^&&&).

A shared lesson across placement and RL is that initialization quality is not exhausted by immediate objective value. In both cases, the initializer is designed to alter the reachable state distribution of subsequent optimization.

6. Recurring principles, trade-offs, and related paradigms

Several recurring design principles appear across these literatures. First, primitive choice is always coupled to a quantitative criterion that controls degeneration. In the barren-plateau setting, the control variable is the contraction factor PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)66^^^^7, with the practical prescription PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)66^^^^8 and widths chosen so that PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)66^^^^9. In identity-block circuits, trainability is preserved by keeping the effective derivative circuit shallow. In truncated diffusion, plausibility is enforced by anchoring at PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6query6^^^^ around a library trajectory. In Gaussian Splatting, assignment is constrained by geometric and color-consistency thresholds. In RL, the central quantity is coverage of reasoning states rather than the accuracy of a single reasoning style (&&&^^^^6query6^^^^&&&, &&&^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^&&&, Xu et al., 5 Apr 2026, Qu et al., 15 Jul 2025, &&&^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^&&&).

Second, mixed-primitive initialization systematically introduces a new model-selection problem. The quantum literature states this most explicitly: avoiding barren plateaus can trade the exponential concentration problem for the challenge of selecting the right trainable pocket among many options. Closely related trade-offs appear elsewhere. Placement reports an initialization gap between computationally expensive area-aware starts and fast point-based starts; the proposed co-optimization smooths the transition between them. Surface reconstruction argues that a single primitive type is insufficient for objects with corners, edges, twigs, and flat patches. Spacecraft ^^^^6query6^^^^DGS shows that primitive initialization is highly effective but sensitive to pose ambiguity. Tailor argues that high chain-of-thought accuracy alone is insufficient if the warm start covers only a narrow reasoning-state distribution (&&&^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^&&&, Qu et al., 15 Jul 2025, Huc et al., 25 Jul 2025, &&&^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6query6^^^^&&&).

Third, mixed-primitive initialization should not be conflated with all forms of structured initialization. A distinct line of work on fractured reservoirs uses a fixed-point, pseudo-time-stepping strategy to reconstruct an unknown equilibrium reference state PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^ consistent with field measurements and geomechanical constraints. There, constitutive laws are written around deviations from PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6max_results6^^^^, and initialization solves

PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6query6^^^^

iteratively until convergence. In the Salt Cove benchmark, the reference state is reported to converge in PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6(Kulshrestha et al., 16 Jun 2026) OR (Grant et al., 2019) OR \6^^^^ iterations, with aperture and flow matching prescribed PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6max_results6 OR \6^^^^ and PRESERVED_PLACEHOLDER_^^^^6Mixed-Primitive-Based Initialization Strategy arXiv (Kulshrestha et al., 16 Jun 2026) barren plateaus identity block initialization (Grant et al., 2019)6^^^^76, while path-dependence tests reveal non-uniqueness near fractures due to Coulomb hysteresis. This is structured initialization, but not primitive mixing in the discrete-compositional sense (&&&^^^^6query6query6^^^^&&&).

A plausible implication is that mixed-primitive-based initialization is best understood as a bias-design methodology. Its function is to reshape the initial measure over states, parameters, or representations so that downstream optimization begins inside a controlled subset of the full search space. The technical challenge then shifts from constructing any nontrivial initializer to selecting, weighting, or learning primitives whose induced bias is matched to the objective, constraints, and geometry of the target problem.