Karp–Sipser Core in Sparse Graphs

• Karp–Sipser core is defined as the residual graph obtained by iteratively removing leaves and their neighbors, and is characterized using fixed-point equations in configuration models.
• In Erdős–Rényi graphs, the core exhibits a phase transition at an average degree of c = e, being negligible below this threshold and scaling linearly above it, with a critical n^(3/5) size at threshold.
• The analysis integrates local weak limits, message-passing recurrences, and differential equation methods, offering insights into matching algorithms, spectral properties, and structural graph decompositions.

to=arxiv_search 大发棋牌json {"query":"Karp-Sipser core configuration model Erdos-Renyi critical core arXiv", "max_results": 10}】【：】【“】【json to=arxiv_search 天天中彩票能json [{"arxiv_id":"(Chatterjee et al., 26 Aug 2025)","title":"Asymptotic size of the Karp-Sipser Core in Configuration Model","authors":"Bingbing Yu, Yung-han Lin","abstract":"We study the asymptotic size of the Karp-Sipser core in the configuration model with arbitrary degree distributions. The Karp-Sipser core is the induced subgraph obtained by iteratively removing all leaves and their neighbors through the leaf-removal process, and finally discarding any isolated vertices \cite{BCC}. Our main result establishes the convergence of the Karp-Sipser core size to an explicit fixed-point equation under general degree assumptions. The approach is based on analyzing the corresponding local weak limit of the configuration model - a unimodular Galton-Watson tree and tracing the evolution process of all vertex states under leaf-removal dynamics by use of the working mechanism of an enhanced version of Warning Propagation along with Node Labeling Propagation.","categories":"math.PR cs.DM","published":"2025-08-26","url":"http://arxiv.org/abs/([2508.19453](/papers/2508.19453))v1"},{"arxiv_id":"([2412.04328](/papers/2412.04328))","title":"The critical Karp--Sipser core of Erdős--Rényi random graphs","authors":"Nicolas Broutin, Benedikt Stufler, Omar Saleh","abstract":"The Karp--Sipser algorithm consists in removing recursively the leaves as well their unique neighbours and all isolated vertices of a given graph. The remaining graph obtained when there is no leaf left is called the Karp--Sipser core. When the underlying graph is the classical sparse Erdős-Rényi random graph $\mathrm{G}[n, \lambda/n]$, it is known to exhibit a phase transition at $\lambda = \mathrm{e}$. We show that at criticality, the Karp--Sipser core has size of order $n^{3/5}$, which proves a conjecture of Bauer and Golinelli. We provide the asymptotic law of this renormalized size as well as a description of the distribution of the core as a graph. Our approach relies on the differential equation method, and builds up on a previous work on a configuration model with bounded degrees.","categories":"math.PR","published":"2024-12-05","url":"http://arxiv.org/abs/([2412.04328](/papers/2412.04328))v1"},{"arxiv_id":"([2212.02463](/papers/2212.02463))","title":"The critical Karp--Sipser core of random graphs","authors":"Alexandre Budzinski, Armand Contat, Nicolas Curien","abstract":"We study the Karp--Sipser core of a random graph made of a configuration model with vertices of degree $1,2$ and $3$. This core is obtained by recursively removing the leaves as well as their unique neighbors in the graph. We settle a conjecture of Bauer & Golinelli and prove that at criticality, the Karp--Sipser core has size $$\approx \mathrm{Cst} \cdot \vartheta^{-2} \cdot n^{3/5}$$ where $\vartheta$ is the hitting time of the curve $t \mapsto \frac{1}{t^{2}}$ by a linear Brownian motion started at $0$. Our proof relies on a detailed multi-scale analysis of the Markov chain associated to Karp-Sipser leaf-removal algorithm close to its extinction time.","categories":"math.PR","published":"2022-12-05","url":"http://arxiv.org/abs/([2212.02463](/papers/2212.02463))v1"},{"arxiv_id":"([2402.05851](/papers/2402.05851))","title":"A central limit theorem for the matching number of a sparse random graph","authors":"Glasgow, Kwan, Sah, Sawhney","abstract":"In 1981, Karp and Sipser proved a law of large numbers for the matching number of a sparse Erdős-Rényi random graph, in an influential paper pioneering the so-called differential equation method for analysis of random graph processes. Strengthening this classical result, and answering a question of Aronson, Frieze and Pittel, we prove a central limit theorem in the same setting: the fluctuations in the matching number of a sparse random graph are asymptotically Gaussian. Our new contribution is to prove this central limit theorem in the subcritical and critical regimes, according to a celebrated algorithmic phase transition first observed by Karp and Sipser. Indeed, in the supercritical regime, a central limit theorem has recently been proved in the PhD thesis of Kreacić, using a stochastic generalisation of the differential equation method (comparing the so-called Karp-Sipser process to a system of stochastic differential equations). Our proof builds on these methods, and introduces new techniques to handle certain degeneracies present in the subcritical and critical cases. Curiously, our new techniques lead to a non-constructive result: we are able to characterise the fluctuations of the matching number around its mean, despite these fluctuations being much smaller than the error terms in our best estimates of the mean. We also prove a central limit theorem for the rank of the adjacency matrix of a sparse random graph.","categories":"math.PR","published":"2024-02-08","url":"http://arxiv.org/abs/([2402.05851](/papers/2402.05851))v1"},{"arxiv_id":"([2105.11718](/papers/2105.11718))","title":"On the Rank, Kernel, and Core of Sparse Random Graphs","authors":"M. Campos, M. Jenssen, M. Michelen, J. Sahasrabudhe","abstract":"We study the rank of the adjacency matrix $A$ of a random Erdos Renyi graph $\lambda = \mathrm{e}$0. It is well known that when $\lambda = \mathrm{e}$1, with high probability, $\lambda = \mathrm{e}$2 is singular. We prove that when $\lambda = \mathrm{e}$3, with high probability, the corank of $\lambda = \mathrm{e}$4 is equal to the number of isolated vertices remaining in $\lambda = \mathrm{e}$5 after the Karp-Sipser leaf-removal process, which removes vertices of degree one and their unique neighbor. We prove a similar result for the random matrix $\lambda = \mathrm{e}$6, where all entries are independent Bernoulli random variables with parameter $\lambda = \mathrm{e}$7. Namely, we show that if $\lambda = \mathrm{e}$8 is the bipartite graph with bi-adjacency matrix $\lambda = \mathrm{e}$9, then the corank of $n^{3/5}$0 is with high probability equal to the max of the number of left isolated vertices and the number of right isolated vertices remaining after the Karp-Sipser leaf-removal process on $n^{3/5}$1. Additionally, we show that with high probability, the $n^{3/5}$2-core of $n^{3/5}$3 is full rank for any $n^{3/5}$4 and $n^{3/5}$5. This partially resolves a conjecture of Van Vu for $n^{3/5}$6. Finally, we give an application of the techniques in this paper to gradient coding, a problem in distributed computing.","categories":"math.PR math.CO","published":"2021-05-25","url":"http://arxiv.org/abs/([2105.11718](/papers/2105.11718))v1"},{"arxiv_id":"([2105.10177](/papers/2105.10177))","title":"Existence of absolutely continuous spectrum for Galton-Watson random trees","authors":"Nicolas Broutin, Abel Fermanian, Jiaoyang Huang, Antoine Maillard","abstract":"We establish a quantitative criterion for an operator defined on a Galton-Watson random tree for having an absolutely continuous spectrum. For the adjacency operator, this criterion requires that the offspring distribution has a relative variance below a threshold. As a byproduct, we prove that the adjacency operator of a supercritical Poisson Galton-Watson tree has a non-trivial absolutely continuous part if the average degree is large enough. We also prove that its Karp and Sipser core has purely absolutely spectrum on an interval if the average degree is large enough. We finally illustrate our criterion on the Anderson model on a d-regular infinite tree with d $n^{3/5}$7 3 and give a quantitative version of Klein's Theorem on the existence of absolutely continuous spectrum at disorder smaller that C $n^{3/5}$8 for some absolute constant C.","categories":"math-ph math.SP math.PR","published":"2021-05-21","url":"http://arxiv.org/abs/([2105.10177](/papers/2105.10177))v1"},{"arxiv_id":"([1311.3254](/papers/1311.3254))","title":"The statistical mechanics of random set packing and a generalization of the Karp-Sipser algorithm","authors":"A. Braunstein, M. Mézard, R. Zecchina","abstract":"We analyse the asymptotic behaviour of random instances of the Maximum Set Packing (MSP) optimization problem, also known as Maximum Matching or Maximum Strong Independent Set on Hypergraphs. We give an analytical prediction of the MSPs size using the 1RSB cavity method from statistical mechanics of disordered systems. We also propose a heuristic algorithm, a generalization of the celebrated Karp-Sipser one, which allows us to rigorously prove that the replica symmetric cavity method prediction is exact for certain problem ensembles and breaks down when a core survives the leaf removal process. The $n^{3/5}$9-phenomena threshold discovered by Karp and Sipser, marking the onset of core emergence and of replica symmetry breaking, is elegantly generalized to $1,2$0 for one of the ensembles considered, where $1,2$1 is the size of the sets.","categories":"cond-mat.dis-nn cs.CC cs.DS math-ph","published":"2013-11-13","url":"http://arxiv.org/abs/([1311.3254](/papers/1311.3254))v1"},{"arxiv_id":"([2403.02140](/papers/2403.02140))","title":"Matching Algorithms in the Sparse Stochastic Block Model","authors":"Aaron Potechin, Yuyang Wang","abstract":"The stochastic block model (SBM) is a generalization of the Erdős-Rényi model of random graphs that describes the interaction of a finite number of distinct communities. In sparse Erdős-Rényi graphs, it is known that a linear-time algorithm of Karp and Sipser achieves near-optimal matching sizes asymptotically almost surely, giving a law-of-large numbers for the matching sizes of such graphs in terms of solutions to an ODE. We provide an extension of this analysis, identifying broad ranges of stochastic block model parameters for which the Karp-Sipser algorithm achieves near-optimal matching sizes, but demonstrating that it cannot perform optimally on general SBM instances. We also consider the problem of constructing a matching online, in which the vertices of one half of a bipartite stochastic block model arrive one-at-a-time, and must be matched as they arrive. We show that the competitive ratio lower bound of 0.837 found by Mastin and Jaillet for the Erdős-Rényi case is tight whenever the expected degrees in all communities are equal. We propose several linear-time algorithms for online matching in the general stochastic block model, but prove that despite very good experimental performance, none of these achieve online asymptotic optimality.","categories":"cs.DS math.CO","published":"2024-03-04","url":"http://arxiv.org/abs/([2403.02140](/papers/2403.02140))v1"},{"arxiv_id":"([1808.10358](/papers/1808.10358))","title":"Asymptotic optimality of degree-greedy discovering of independent sets in Configuration Model graphs","authors":"Armand Bordenave, Marc Lelarge, Justin Salez","abstract":"Finding independent sets of maximum size in fixed graphs is well known to be an NP-hard task. Using scaling limits, we characterise the asymptotics of sequential degree-greedy explorations and provide sufficient conditions for this algorithm to find an independent set of asymptotically optimal size in large sparse random graphs with given degree sequences. In the special case of sparse Erdös-Rényi graphs, our results allow to give a simple proof of the so-called $1,2$2-phenomenon identified by Karp and Sipser for matchings and to give an alternative characterisation of the asymptotic independence number.","categories":"math.PR","published":"2018-08-30","url":"http://arxiv.org/abs/([1808.10358](/papers/1808.10358))v2"},{"arxiv_id":"([1811.11035](/papers/1811.11035))","title":"Finding perfect matchings in random regular graphs in linear time","authors":"Markus Anastos, Alan Frieze","abstract":"In a seminal paper on finding large matchings in sparse random graphs, Karp and Sipser proposed two algorithms for this task. The second algorithm has been intensely studied, but due to technical difficulties, the first algorithm has received less attention. Empirical results suggest that the first algorithm is superior. In this paper we analyze an adapted version of the first algorithm, the Reduce-Construct algorithm. We show that the Reduce-Construct algorithm finds a maximum matching in random $1,2$3-regular graphs in linear time in expectation, as opposed to $1,2$4 time for the worst-case.","categories":"math.CO","published":"2018-11-27","url":"http://arxiv.org/abs/([1811.11035](/papers/1811.11035))v1"}] The Karp–Sipser core is the residual graph produced by the leaf-removal phase of the Karp–Sipser matching algorithm: one iteratively removes every degree-1 vertex together with its unique neighbor, discards isolated vertices created along the way, and stops when no leaves remain. In recent configuration-model work, the core is treated as the induced subgraph on the surviving vertices, with asymptotic size characterized by an explicit fixed-point system derived from local weak limits and message passing (Chatterjee et al., 26 Aug 2025). In sparse Erdős–Rényi graphs, its emergence exhibits the classical threshold at average degree $1,2$5, while at criticality its size is neither linear nor tight but of order $1,2$6 with a Brownian-hitting-time limit law (Budzinski et al., 2024).

1. Definition, leaf-removal dynamics, and the 2-core distinction

The Karp–Sipser leaf-removal process proceeds by repeatedly identifying leaves, removing each leaf together with its unique neighbor, deleting incident edges, and discarding isolated vertices. The resulting leaf-free residual graph is the Karp–Sipser core. This object is intrinsic in the sense that the terminal residual does not depend on the order of leaf removals; the Abelian property is emphasized in critical-core analyses (Budzinski et al., 2022).

A central technical point is that the Karp–Sipser core is not, in general, the standard 2-core. The standard 2-core is obtained by iteratively removing only vertices of degree $1,2$7 or $1,2$8, whereas Karp–Sipser removes a degree-1 vertex and its neighbor simultaneously. Because a neighbor of a leaf may have degree larger than $1,2$9, Karp–Sipser pruning is strictly more aggressive, and the residual can be strictly smaller than the 2-core (Chatterjee et al., 26 Aug 2025). This distinction is also made explicit in analyses of matching-number fluctuations and degree-greedy dynamics, where the Karp–Sipser core is described as contained in, and potentially strictly smaller than, the 2-core (Glasgow et al., 2024, Jonckheere et al., 2018).

Some papers in adjacent contexts use “KS core” more loosely for the residual minimum-degree-at-least-2 graph and identify it with the 2-core in random graph terminology or on simple graphs (DeMichele et al., 2021, Brandenberger et al., 2024). This reflects a terminological divergence rather than a uniform definition. The more precise recent random-graph analyses distinguish the Karp–Sipser core from the classical 2-core whenever the simultaneous deletion of neighbors matters.

The residual graph has minimum degree at least $3$0, but that condition alone does not characterize it. Its significance comes from algorithmics: phase I of Karp–Sipser performs all leaf-based “safe” matching moves, and phase II operates on the residual core.

2. Configuration-model asymptotics and fixed-point characterization

For the configuration model with empirical degree distribution converging to $3$1, finite mean

$3$2

and generating functions

$3$3

the asymptotic Karp–Sipser core fraction is described by a nested fixed-point system (Chatterjee et al., 26 Aug 2025). Writing $3$4, one defines

$3$5

Let $3$6 and $3$7 denote the smallest and largest solutions of $3$8 in $3$9. Under weak convergence of degrees, $$\approx \mathrm{Cst} \cdot \vartheta^{-2} \cdot n^{3/5}$$0, the Molloy–Reed-type condition

$$\approx \mathrm{Cst} \cdot \vartheta^{-2} \cdot n^{3/5}$$1

and a stability condition

$$\approx \mathrm{Cst} \cdot \vartheta^{-2} \cdot n^{3/5}$$2

the normalized core size converges in probability to

$$\approx \mathrm{Cst} \cdot \vartheta^{-2} \cdot n^{3/5}$$3

This gives an explicit asymptotic formula for arbitrary degree distributions within the stated regime (Chatterjee et al., 26 Aug 2025).

The derivation uses the local weak limit of the configuration model, namely a unimodular Galton–Watson tree, together with an enhanced Warning Propagation recursion on a three-symbol alphabet $$\approx \mathrm{Cst} \cdot \vartheta^{-2} \cdot n^{3/5}$$4. In this formulation, red means “leaf,” blue means “neighbor-of-leaf,” and green means “candidate core.” If child-to-parent messages have law $$\approx \mathrm{Cst} \cdot \vartheta^{-2} \cdot n^{3/5}$$5, then the outgoing message law is

$$\approx \mathrm{Cst} \cdot \vartheta^{-2} \cdot n^{3/5}$$6

Starting from $$\approx \mathrm{Cst} \cdot \vartheta^{-2} \cdot n^{3/5}$$7, $$\approx \mathrm{Cst} \cdot \vartheta^{-2} \cdot n^{3/5}$$8, the iteration converges to

$$\approx \mathrm{Cst} \cdot \vartheta^{-2} \cdot n^{3/5}$$9

and Node Labeling Propagation then yields the root-survival probability

$\vartheta$0

which reduces to the stated formula after substituting $\vartheta$1 (Chatterjee et al., 26 Aug 2025).

This framework is notable because it replaces differential-equation arguments by a unimodular-tree recursion and produces a computable description for arbitrary degree laws. A plausible implication is that the Karp–Sipser core can now be analyzed in heterogeneous sparse ensembles using the same fixed-point machinery already familiar from local weak convergence and cavity methods.

3. Erdős–Rényi phase transition and critical scaling

In $\vartheta$2 with $\vartheta$3, the degree law is Poisson$\vartheta$4, so

$\vartheta$5

The configuration-model fixed-point equation becomes

$\vartheta$6

and the limiting core fraction is

$\vartheta$7

(Chatterjee et al., 26 Aug 2025). The emergence threshold is $\vartheta$8: below $\vartheta$9, the Karp–Sipser core vanishes or is negligible, while for $t \mapsto \frac{1}{t^{2}}$0 one has separated solutions $t \mapsto \frac{1}{t^{2}}$1 and $t \mapsto \frac{1}{t^{2}}$2 (Chatterjee et al., 26 Aug 2025).

The $t \mapsto \frac{1}{t^{2}}$3-phenomenon is also recovered through alternative fluid-limit and degree-greedy analyses. In sparse Erdős–Rényi graphs, the Karp–Sipser core is asymptotically empty for $t \mapsto \frac{1}{t^{2}}$4 and has linear size for $t \mapsto \frac{1}{t^{2}}$5; by contrast, the 2-core already appears at $t \mapsto \frac{1}{t^{2}}$6, which sharply illustrates the difference between the two notions (Jonckheere et al., 2018). Matching-number analyses formulate the same phase transition as

$t \mapsto \frac{1}{t^{2}}$7

with $t \mapsto \frac{1}{t^{2}}$8 for $t \mapsto \frac{1}{t^{2}}$9 and $0$0 for $0$1 (Glasgow et al., 2024).

At the exact critical point, the Karp–Sipser core has anomalous scale $0$2. For the configuration model with degrees in $0$3, Budzinski, Contat, and Curien prove that criticality is governed by

$0$4

They show that $0$5 is $0$6 in the subcritical regime $0$7, has linear size with

$0$8

in the supercritical regime $0$9, and at $A$0 satisfies

$A$1

where

$A$2

for standard Brownian motion $A$3 (Budzinski et al., 2022).

For the critical Erdős–Rényi graph $A$4, the same $A$5 scaling is established with a considerably finer structural description. If $A$6 is the number of degree-$A$7 vertices in the critical Karp–Sipser core, then jointly

$A$8

converges in distribution to an explicit vector involving powers of the same Brownian hitting time $A$9, and $\lambda = \mathrm{e}$00 with high probability for all $\lambda = \mathrm{e}$01 (Budzinski et al., 2024). Conditionally on the degree counts, the critical core is a configuration model conditioned to be simple (Budzinski et al., 2024). This identifies not only the scaling exponent but also the internal degree structure of the critical residue.

4. Role in matching algorithms and algorithmic optimality

The Karp–Sipser core measures precisely what phase I of the Karp–Sipser algorithm fails to eliminate. Phase I greedily matches leaves to their unique neighbors and deletes both endpoints; these moves are “safe” in the sense that they can be extended to a maximum matching (Brandenberger et al., 2024). The residual core is therefore the algorithmically difficult part of the instance.

This relation is explicit in Erdős–Rényi graphs. When $\lambda = \mathrm{e}$02, phase I is essentially optimal: the Karp–Sipser core vanishes or is negligible, and the residual contributes only $\lambda = \mathrm{e}$03 to the maximum matching (Chatterjee et al., 26 Aug 2025). In the supercritical regime $\lambda = \mathrm{e}$04, a linear core remains, and phase II must match the residual graph; nevertheless, in Erdős–Rényi graphs the gap between the phase-I matching and the maximum matching remains sublinear (Chatterjee et al., 26 Aug 2025). The law of large numbers and central limit theorem for the matching number are built around the decomposition

$\lambda = \mathrm{e}$05

where $\lambda = \mathrm{e}$06 is the number of leaf-removal steps and $\lambda = \mathrm{e}$07 is the Karp–Sipser core (Glasgow et al., 2024).

Fluid-limit analyses make this connection quantitative. In the classical sparse regime $\lambda = \mathrm{e}$08, the maximum matching size satisfies

$\lambda = \mathrm{e}$09

with

$\lambda = \mathrm{e}$10

(Glasgow et al., 2024). In equitable stochastic block models, including Erdős–Rényi as a special case, the asymptotic maximum matching size is

$\lambda = \mathrm{e}$11

and Karp–Sipser achieves this up to $\lambda = \mathrm{e}$12 (Brandenberger et al., 2024).

In sparse stochastic block models, the size and structure of the Karp–Sipser core cease to be a one-parameter phenomenon. A sufficient condition for subcriticality is

$\lambda = \mathrm{e}$13

which implies that the Karp–Sipser core has size $\lambda = \mathrm{e}$14 with high probability (Brandenberger et al., 2024). More generally, a multitype fixed-point system controls whether leaf-removal annihilates the graph: $\lambda = \mathrm{e}$15 Uniqueness of the solution on $\lambda = \mathrm{e}$16 implies subcriticality of leaf removal and hence an $\lambda = \mathrm{e}$17 core (Brandenberger et al., 2024). Conversely, explicit counterexamples show that a large, heterogeneous Karp–Sipser core can make phase II far from optimal, even when a near-perfect matching exists (Brandenberger et al., 2024).

A recurrent misconception is that only the size of the core matters. The SBM counterexamples suggest that its internal heterogeneity also matters: when the core is $\lambda = \mathrm{e}$18 and structurally uneven, uniform random edge choices in phase II can waste matching capacity in dense regions while leaving sparse regions uncovered.

5. Linear-algebraic and spectral connections

The Karp–Sipser core also controls structural linear algebra in sparse random graphs. For $\lambda = \mathrm{e}$19, the adjacency matrix of $\lambda = \mathrm{e}$20 with $\lambda = \mathrm{e}$21, the corank is, with high probability, exactly the number of isolated vertices remaining after Karp–Sipser leaf removal: $\lambda = \mathrm{e}$22 and the adjacency matrix of the Karp–Sipser core is full rank: $\lambda = \mathrm{e}$23 (DeMichele et al., 2021). In the bipartite Bernoulli case, the corresponding statement is

$\lambda = \mathrm{e}$24

with the KS-core bi-adjacency matrix having full row rank or full column rank with high probability (DeMichele et al., 2021).

The mechanism behind this result is the invariance of corank under leaf removal together with the characterization of minimal linear dependencies as tree dependencies with high probability (DeMichele et al., 2021). In effect, leaf-removal peels exactly the tree-like structures that support kernel vectors, leaving a full-rank core. This places the Karp–Sipser core at the intersection of combinatorial peeling and sparse random matrix theory.

On Galton–Watson trees, the Karp–Sipser core also appears in spectral theory. For a Poisson Galton–Watson tree with offspring distribution $\lambda = \mathrm{e}$25, there is a threshold $\lambda = \mathrm{e}$26: if $\lambda = \mathrm{e}$27, the core is empty almost surely, whereas for $\lambda = \mathrm{e}$28 the core is non-empty almost surely (Arras et al., 2021). Conditioned on the root lying in the core, the connected component of the root in the core has law $\lambda = \mathrm{e}$29, where $\lambda = \mathrm{e}$30 is $\lambda = \mathrm{e}$31 conditioned on being at least $\lambda = \mathrm{e}$32, with $\lambda = \mathrm{e}$33 for $\lambda = \mathrm{e}$34 and $\lambda = \mathrm{e}$35 as $\lambda = \mathrm{e}$36 (Arras et al., 2021).

For large average degree, the adjacency operator on this leaf-free core has purely absolutely continuous spectrum on a macroscopic interval: for any $\lambda = \mathrm{e}$37, if $\lambda = \mathrm{e}$38 is large enough, the spectral measure is absolutely continuous with almost-everywhere positive density on

$\lambda = \mathrm{e}$39

(Arras et al., 2021). The paper’s interpretation is that removing leaves eliminates dense atomic contributions from finite pendant trees, so the Karp–Sipser core isolates the extended spectral component (Arras et al., 2021).

6. Generalizations, critical phenomena, and broader variants

The Karp–Sipser core has natural extensions beyond classical sparse graphs. In random set packing and maximum strong independent set on hypergraphs, Braunstein, Mézard, and Zecchina define a generalized Karp–Sipser algorithm on bipartite factor graphs and identify the analogue of leaf-removal through “pendants” (Lucibello et al., 2013). In the $\lambda = \mathrm{e}$40-uniform ensemble $\lambda = \mathrm{e}$41, the generalized core emerges at the threshold

$\lambda = \mathrm{e}$42

which reduces to the classical $\lambda = \mathrm{e}$43-threshold when $\lambda = \mathrm{e}$44 (Lucibello et al., 2013). In that setting, survival of the core coincides with breakdown of replica symmetry and onset of a nontrivial 1RSB regime (Lucibello et al., 2013).

The same paper formulates leaf-removal through the two-variable system

$\lambda = \mathrm{e}$45

where

$\lambda = \mathrm{e}$46

When the smallest solution is unique and satisfies $\lambda = \mathrm{e}$47, the core density vanishes; when a second solution with $\lambda = \mathrm{e}$48 appears, a positive core survives (Lucibello et al., 2013). This provides a statistical-mechanics interpretation of the Karp–Sipser core as an order parameter for the onset of clustered solution geometry.

There are also model-specific adaptations in which the “core” changes because leaves are absent initially. In random $\lambda = \mathrm{e}$49-regular graphs with fixed $\lambda = \mathrm{e}$50, Anastos and Frieze analyze the Reduce–Construct algorithm, which creates degree-1 and degree-2 vertices through controlled edge deletions and contractions. The evolving residual graph has minimum degree at least $\lambda = \mathrm{e}$51 and functions as a dynamic “KS 3-core sequence” rather than the classical 2-core (Anastos, 2018). This adaptation is natural in that model, but it should not be conflated with the standard Karp–Sipser core of sparse graphs with intrinsic leaves.

Across these variants, one theme is consistent. The Karp–Sipser core is the obstruction left after all leaf-driven reductions have been exhausted. Its asymptotic size can be negligible, critical, or linear; its internal structure can control matching performance, nullity, and spectral type; and in modern treatments it is increasingly described by explicit fixed points, stochastic differential limits, or message-passing recursions rather than only by combinatorial peeling arguments.