Matroid Equitability Conjecture Resolution

• Matroid Equitability Conjecture is a balancing problem where a matroid's ground set is partitioned into bases such that any subset's intersections differ by at most one.
• It employs refined exchange theory and directed bipartite exchange graphs to systematically redistribute elements and achieve equitable partitions.
• The algorithmic solution has significant implications for fair division, enabling robust EF1 and MMS allocations under matroid constraints.

The Matroid Equitability Conjecture concerns the extent to which a partition of a matroid ground set into bases can be made balanced with respect to a prescribed subset. In its original form, posed by Fekete and Szabó in 2011, it asked whether every matroid whose ground set can be partitioned into two bases admits, for every subset $S \subseteq E$, a decomposition $E=B_1 \sqcup B_2$ into two bases such that $||B_1\cap S|-|B_2\cap S||\le 1$. Akrami, Raj, and Végh proved a substantially stronger statement: if the ground set can be partitioned into $k \ge 1$ disjoint bases, then for every $S\subseteq E$ there is a partition into $k$ bases whose intersections with $S$ all have sizes between $\lfloor |S|/k\rfloor$ and $\lceil |S|/k\rceil$, and such a partition can be found in polynomial time (Akrami et al., 16 Jul 2025). The same work also establishes near-equitable splittings for two disjoint sets and derives applications to matroid-constrained fair division (Akrami et al., 16 Jul 2025).

1. Statement of the conjecture and its resolution

Let $M=(E,\mathcal I)$ be a matroid of rank $E=B_1 \sqcup B_2$0. In the formulation highlighted by Fekete and Szabó, one assumes that the ground set $E=B_1 \sqcup B_2$1 can be partitioned into two bases. The matroid is called equitable if, for every subset $E=B_1 \sqcup B_2$2, there exists a partition $E=B_1 \sqcup B_2$3 into two bases satisfying

$E=B_1 \sqcup B_2$4

The question “Is every matroid equitable?” was left open in Electron. J. Comb. 2011 and became known as the matroid equitability conjecture.

The theorem proved in "Matroids are Equitable" (Akrami et al., 16 Jul 2025) subsumes this original two-base problem. If $E=B_1 \sqcup B_2$5 can be partitioned into $E=B_1 \sqcup B_2$6 disjoint bases, then for every $E=B_1 \sqcup B_2$7 there exists a partition

$E=B_1 \sqcup B_2$8

into $E=B_1 \sqcup B_2$9 bases such that for all $||B_1\cap S|-|B_2\cap S||\le 1$0,

$||B_1\cap S|-|B_2\cap S||\le 1$1

equivalently,

$||B_1\cap S|-|B_2\cap S||\le 1$2

This settles the conjecture in the affirmative and, in the one-set setting, achieves the strongest balancing compatible with integrality (Akrami et al., 16 Jul 2025).

The result is algorithmic as well as existential: the equitable partition can be found in polynomial time. This algorithmic aspect is integral to the theorem rather than a secondary consequence.

2. Historical and conceptual context

The equitability problem sits at the interface of basis exchange theory, fair representation, and combinatorial balancing. In a single matroid, fair representation phenomena are classical. Aharoni, Berger, Kotlar, and Ziv formulate this in terms of a simplicial complex parameter $||B_1\cap S|-|B_2\cap S||\le 1$3, the minimum size of an edge-cover, and recall that when $||B_1\cap S|-|B_2\cap S||\le 1$4 is a matroid, Edmonds’ theorem gives $||B_1\cap S|-|B_2\cap S||\le 1$5 (Aharoni et al., 2016). They then state the folklore/Edmonds fair-representation theorem: if $||B_1\cap S|-|B_2\cap S||\le 1$6 is a matroid and $||B_1\cap S|-|B_2\cap S||\le 1$7, then for every partition $||B_1\cap S|-|B_2\cap S||\le 1$8 there exists an independent set $||B_1\cap S|-|B_2\cap S||\le 1$9 with

$k \ge 1$0

for each $k \ge 1$1 (Aharoni et al., 2016).

The matroid equitability problem differs in a decisive way. Rather than selecting a single independent set that represents each part fairly, it seeks a partition of the entire ground set into bases with all basis-intersections with a prescribed subset as equal as possible. This makes the problem inherently reconfiguration-based: one is not merely proving existence of a favorable basis, but existence of a globally balanced basis decomposition.

The work on fair representation in the intersection of two matroids provides a further contextual link. For a dimatroid $k \ge 1$2, Aharoni–Berger–Kotlar–Ziv conjecture almost-fair representation bounds and prove the two-part case using truncation, fractional covering, and exchange-sequences in the dimatroid (Aharoni et al., 2016). The later theorem that matroids are equitable is described as settling several special-case fair-representation conjectures for the matroid/dual-matroid pair (Akrami et al., 16 Jul 2025). This suggests that equitability, though weaker than the full two-matroid representation problem, captures a structurally central case.

The same later paper also records conceptual links to classical exchange conjectures: White’s and Gabow’s basis-sequence conjectures would imply equitability as a corollary, while equitability is strictly weaker (Akrami et al., 16 Jul 2025). Accordingly, the conjecture belongs to the broader program of understanding how far basis exchange can be pushed toward canonical balancing statements.

3. Exchange structures underlying the proof

The proof of equitability is organized around a refined exchange theory for pairs of disjoint bases (Akrami et al., 16 Jul 2025). Let $k \ge 1$3 be disjoint bases. A set $k \ge 1$4 is called exchangeable if both $k \ge 1$5 and $k \ge 1$6 are bases. Given a subset $k \ge 1$7 and an element $k \ge 1$8, one says that $k \ge 1$9 is $S\subseteq E$0-exchangeable if

$S\subseteq E$1

This notion packages precisely the exchanges needed to move one unit of $S\subseteq E$2-mass from one basis to another while preserving the basis property.

The corresponding combinatorial object is the directed bipartite exchange graph $S\subseteq E$3 on $S\subseteq E$4. Its edges are

$S\subseteq E$5

Here $S\subseteq E$6 denotes the family of bases. The graph encodes admissible one-element transfers between the two bases.

Several standard and nonstandard exchange principles are then deployed. The symmetric exchange lemma states that for each $S\subseteq E$7 there exists $S\subseteq E$8 with mutual exchange. More generally, Schrijver’s matching-exchange statement implies that any perfect matching in the induced bipartite graph $S\subseteq E$9 corresponds to an exchangeable set (Akrami et al., 16 Jul 2025). Directed cycles in the exchange graph are especially useful: any directed cycle $k$0 yields an exchangeable set $k$1, and a chordless cycle gives a unique perfect matching in each bipartite half, so by the matroid-matching theorem $k$2 is exchangeable (Akrami et al., 16 Jul 2025).

When immediate exchanges are unavailable, the argument turns to a more delicate circuit analysis. If one cannot directly find a short cycle or symmetric exchange between $k$3 and $k$4, one studies a strongly connected component $k$5 of $k$6 with $k$7. One then constructs a family $k$8 of fundamental-type circuits satisfying specified witness properties, and applies strong circuit-exchange to combine them into a single circuit with a contradictory exchange behavior, thereby forcing the desired cycle (Akrami et al., 16 Jul 2025). This circuit-family argument is the technical core of the proof.

4. The exchange theorem and the polynomial-time balancing algorithm

The global balancing theorem is obtained by iterating a two-base exchange statement. The crucial result is Theorem 2.2 of "Matroids are Equitable" (Akrami et al., 16 Jul 2025):

Let $k$9 be disjoint bases with $S$0. Then there exists $S$1 and an associated $S$2-exchangeable set $S$3, computable in polynomial time.

Its proof proceeds through the exchange graph. If there is a direct symmetric edge from some $S$4 to some $S$5, then $S$6 already works. Otherwise, the strongly connected component and circuit-family machinery is used to produce a directed cycle $S$7 with exactly one vertex $S$8; taking a minimal such cycle yields a chordless cycle and hence an exchangeable set $S$9 (Akrami et al., 16 Jul 2025).

From this theorem, the balancing algorithm is straightforward in outline. Start from any partition of $\lfloor |S|/k\rfloor$0 into $\lfloor |S|/k\rfloor$1 bases. If there are indices $\lfloor |S|/k\rfloor$2 with

$\lfloor |S|/k\rfloor$3

focus on the pair $\lfloor |S|/k\rfloor$4. The exchange theorem provides a $\lfloor |S|/k\rfloor$5-exchangeable set $\lfloor |S|/k\rfloor$6 that swaps one element $\lfloor |S|/k\rfloor$7 from $\lfloor |S|/k\rfloor$8 with some element of $\lfloor |S|/k\rfloor$9. Replacing $\lceil |S|/k\rceil$0 by $\lceil |S|/k\rceil$1 increases $\lceil |S|/k\rceil$2 by $\lceil |S|/k\rceil$3 without affecting the other bases (Akrami et al., 16 Jul 2025).

Each step decreases the range

$\lceil |S|/k\rceil$4

by at least $\lceil |S|/k\rceil$5. Consequently, within $\lceil |S|/k\rceil$6 steps one reaches a configuration in which all intersection sizes differ by at most one (Akrami et al., 16 Jul 2025). Since each step is polynomial-time computable, the entire procedure is polynomial-time. The theorem is therefore both structural and constructive.

5. Two-set near-equitability

The paper extends the one-set theorem to the simultaneous balancing of two disjoint subsets $\lceil |S|/k\rceil$7 (Akrami et al., 16 Jul 2025). Exact $\lceil |S|/k\rceil$8 balancing in both coordinates cannot be guaranteed for arbitrary matroids: the $\lceil |S|/k\rceil$9 graphic matroid gives an example in which an unavoidable $M=(E,\mathcal I)$0 discrepancy appears in one of the two sets (Akrami et al., 16 Jul 2025). The obstruction is therefore genuine rather than an artifact of the proof.

Under the same hypothesis that $M=(E,\mathcal I)$1 can be partitioned into $M=(E,\mathcal I)$2 bases, Theorem 3.1 shows that there exists a $M=(E,\mathcal I)$3-base partition such that:

1. for all $M=(E,\mathcal I)$4,

$M=(E,\mathcal I)$5

2. for all $M=(E,\mathcal I)$6,

$M=(E,\mathcal I)$7

3. a similar bound holds for $M=(E,\mathcal I)$8.

For $M=(E,\mathcal I)$9, Corollary 3.2 yields a parity-sensitive trichotomy (Akrami et al., 16 Jul 2025). The possibilities are summarized below.

Parity of $E=B_1 \sqcup B_2$00	Conclusion for two bases
both odd	exact partition of each
one odd, one even	exact on the odd, difference $E=B_1 \sqcup B_2$01 on the even
both even	one set splits exactly, the other has difference at most $E=B_1 \sqcup B_2$02

The paper describes this as tight. In particular, the appearance of the bound $E=B_1 \sqcup B_2$03 in the two-set setting is not merely technical. A plausible implication is that the transition from one monitored subset to two monitored subsets changes the balancing problem qualitatively: the one-set case is governed by exact floor/ceiling balancing, while the two-set case is constrained by unavoidable interaction effects between the two coordinates.

6. Applications to matroid-constrained fair division

The same exchange and balancing results yield two applications in fair division under a common matroid constraint on the item set $E=B_1 \sqcup B_2$04, where every allocated bundle must be a basis and valuations are additive (Akrami et al., 16 Jul 2025).

The first concerns identical tri-valued valuations. If all agents share the same additive valuation $E=B_1 \sqcup B_2$05 taking exactly three values $E=B_1 \sqcup B_2$06 with $E=B_1 \sqcup B_2$07, then by shifting one may assume the values are $E=B_1 \sqcup B_2$08. The two-set near-equitability theorem is applied to the set of $E=B_1 \sqcup B_2$09-items and the set of $E=B_1 \sqcup B_2$10-items, producing basis bundles in which the high-value items are split within $E=B_1 \sqcup B_2$11 and the low-value items within $E=B_1 \sqcup B_2$12. A subsequent local exchange argument reduces the remaining envy to envy-free up to one item, establishing the existence of a matroid-constrained EF1 allocation (Theorem 4.2) (Akrami et al., 16 Jul 2025).

The second concerns bi-valued additive valuations. If each valuation satisfies $E=B_1 \sqcup B_2$13, with $E=B_1 \sqcup B_2$14 possibly agent-dependent, then Theorem 1.1 implies that agent $E=B_1 \sqcup B_2$15’s max-min-share satisfies

$E=B_1 \sqcup B_2$16

and is exactly achievable by splitting $E=B_1 \sqcup B_2$17 into $E=B_1 \sqcup B_2$18 bases (Corollary 4.3) (Akrami et al., 16 Jul 2025). A standard lone-divider / Hall-matching induction then allocates these bases one at a time so that each agent receives at least her $E=B_1 \sqcup B_2$19. This yields a complete MMS-fair allocation (Theorem 4.4) (Akrami et al., 16 Jul 2025).

These applications show that the equitability theorem is not only a basis-exchange result but also a fairness mechanism under combinatorial feasibility constraints. The fair-division consequences are existence results, and in the second case they are exact rather than approximate.

7. Relations, corollaries, and open directions

Several further consequences and research directions are recorded in the paper (Akrami et al., 16 Jul 2025). The polynomial-time algorithm for equitable splitting into $E=B_1 \sqcup B_2$20 bases gives a deterministic algorithm for Exact Matroid Intersection in the special self-dual case $E=B_1 \sqcup B_2$21. The result also settles several special-case fair-representation conjectures of Aharoni–Berger–Kotlar–Ziv for the matroid/dual-matroid pair.

An explicit broader conjectural program is proposed. For $E=B_1 \sqcup B_2$22 disjoint sets $E=B_1 \sqcup B_2$23, one asks whether there is a universal bound $E=B_1 \sqcup B_2$24 such that every $E=B_1 \sqcup B_2$25 can be kept within $E=B_1 \sqcup B_2$26 of its mean $E=B_1 \sqcup B_2$27. The paper settles the cases $E=B_1 \sqcup B_2$28 and $E=B_1 \sqcup B_2$29, with $E=B_1 \sqcup B_2$30, and asks whether $E=B_1 \sqcup B_2$31 might hold for all $E=B_1 \sqcup B_2$32 (Akrami et al., 16 Jul 2025). This is a natural extension of the one-set and two-set theorems, though no general answer is provided.

These directions connect back to earlier work on intersections of matroids. In the dimatroid setting, Aharoni–Berger–Kotlar–Ziv prove a two-part almost-fair representation theorem with parameter

$E=B_1 \sqcup B_2$33

and formulate open problems including the cases of three or more parts, removal of the $E=B_1 \sqcup B_2$34 penalty, and stronger exchange lemmas (Aharoni et al., 2016). The equitability theorem for matroids does not resolve those dimatroid questions, but it clarifies that in the single-matroid case the exact balancing target is attainable.

In aggregate, the resolution of the Matroid Equitability Conjecture establishes that basis partitions can always be reorganized to equidistribute any prescribed subset as evenly as integrality permits, extends this to a sharp two-set near-equitability statement, and connects the resulting exchange framework to fair representation, exact matroid-intersection phenomena in a special case, and matroid-constrained EF1 and MMS guarantees (Akrami et al., 16 Jul 2025).