Research
Boost+: Equitable, Incentive-Compatible Block Building
Ethereum's current block-building pipeline, MEV-Boost, has become centralized because searchers often send profitable bundles exclusively to a single builder ("integration"), which distorts…

- arXiv
- 2602.04007
- Published
- 2026-02-03
- Authors
- Mengqian Zhang, Sen Yang, Kartik Nayak, Fan Zhang
AI summary
Overview
- Research area: Blockchain systems security and mechanism design (cs.CR), specifically Ethereum block building, proposer–builder separation (PBS), and maximal extractable value (MEV).
- Technical level: Advanced. The paper combines game theory (VCG mechanism design, dominant-strategy incentive compatibility) with systems design (TEEs, RPC services, algorithm execution in 12-second slots), though the core idea and the empirical findings are accessible to a motivated non-specialist.
- Scope: The paper proposes Boost+, a block-building architecture and accompanying mechanism that aims to remove the incentive for searchers and builders to form exclusive "integrations," while keeping block building computationally feasible within an Ethereum slot.
What This Paper Is About
Ethereum's current block-building pipeline, MEV-Boost, has become centralized because searchers often send profitable bundles exclusively to a single builder ("integration"), which distorts competition, wastes scarce block space when compatible transactions from different builders are never combined, and hides MEV flows through opaque off-chain payments. The paper asks whether the incentive to integrate can be eliminated through mechanism and system design rather than through policy or voluntary reform. It answers with Boost+, a design that separates transaction collection from algorithm execution and governs searchers, builders, and proposers together inside a trusted execution environment.
Key Contributions
- Boost+, a three-party block-building architecture. Unlike MEV-Boost, which does not treat searchers as participants in the mechanism design, Boost+ explicitly models all three practically important actors (searchers, builders, and proposers) and decouples the process into collecting transactions/order flows and executing block-building algorithms against them.
- A mechanism, denoted M_Boost+. It is built on a VCG-style framework and gives a two-stage structure: first the default algorithm computes a default block and per-order-flow refunds; then builder-submitted algorithms run to determine the winning block and the value split among builders and the proposer. Builders who win receive the additional surplus over the second-highest block, so good algorithms — not private partnerships — are rewarded.
- Formal incentive guarantees. Truthful bidding is a dominant strategy for all builders. For searchers, truthful reporting is dominant whenever the default algorithm dominates competing builders, and remains dominant for all conflict-free transactions even when builders may win. For conflict-free order flows, choosing not to integrate plus truthful bidding still dominates any joint deviation, even when a searcher is technically capable of integrating with a builder.
- A concrete default algorithm grounded in empirical study. The authors analyze real Ethereum transaction conflicts, find that conflict groups are typically small, and use that to build an algorithm that groups transactions by read/write conflicts and enumerates within groups. It constructs the optimal block in 53.2% of cases in their evaluation on historical transaction data.
Main Findings
- Integration is the centralization driver. Integration — a searcher sending profitable bundles exclusively to one builder — gives that builder an unfair advantage in the MEV-Boost auction. The authors report that as of August 2025, two builders produce the majority of blocks, and cite the Banana Gun–Titan pairing as a well-known integration pattern.
- Integration also wastes block space. Different integrated builders often hold valuable transactions from different searchers that do not conflict, yet only one builder's block is chosen, so many compatible transactions are excluded even though they could have been included together.
- Naive VCG is truthful but infeasible. With an idealized oracle that can maximize over all feasible blocks, direct VCG gives dominant-strategy incentive compatibility (DSIC) for searchers and welfare maximization (Theorem 1). But full block space grows exponentially with the number of bundles, and welfare-maximizing block construction subsumes NP-hard problems, so exact implementation is intractable; a slot is roughly 12 seconds on Ethereum.
- A tractable VCG variant preserves truthfulness. Replacing the oracle with any algorithm A that (i) considers a set of blocks independent of bids and (ii) can find the best block and each per-bundle counterfactual block in polynomial time yields a mechanism that is efficient and DSIC for searchers (Theorem 2). Welfare loss relative to the optimum depends on how good A is.
- Conflict-free order flows get zero net payment. The mechanism separates out transactions that do not touch storage read/written by others, always appends them to the winning block, and refunds their bid in full.
- Empirical conflict structure is favorable. Grouping Ethereum transactions by storage conflicts shows that in most blocks conflict groups are small, with only a few groups containing multiple transactions — consistent with MEV opportunities being limited in most blocks. Most groups are small enough for full enumeration; special cases such as sandwich attacks need only a single pass.
- Integration cannot be fully eliminated, only made harder. A strategic searcher can still simulate integration by making a bundle execute only when
block.coinbaseequals a target address, reverting otherwise. To non-integrating builders the bundle behaves as a no-op. Hidingblock.coinbasecan be bypassed via state-dependent triggering; the achievable goal is that a bundle can identify its integrated algorithm but cannot distinguish the default algorithm from other algorithms. - Builder isolation blunts reactive attacks. Builder-submitted algorithms run in isolated TEEs and cannot alter transactions, inspect other algorithms, or communicate during execution, so sandwich and imitation attacks that depend on observing a searcher's bundle in real time are reduced to blind pre-submission guesses.
- Default algorithm quality. The concrete default algorithm is reported optimal in 53.2% of cases; the paper content available here does not report further evaluation metrics such as average welfare loss, runtime distributions, or comparison against specific real builders.
Methodology in Plain English
The authors start from a diagnosis: today's pipeline lets a searcher and a builder strike a private deal, and the system has no way to police that deal because searchers are outside the mechanism. So they redesign the pipeline so that everyone submits into one governed process.
In Boost+, searchers (and ordinary users) submit their transactions — "order flows" — with the amount they are willing to pay, through an RPC service inside a TEE. Builders submit algorithms, not finished blocks. The mechanism then runs every submitted algorithm plus its own default algorithm over the same pool of transactions, picks the candidate block whose total payment is highest, and computes refunds.
To get truthful bidding, they borrow the VCG idea: each participant effectively pays the harm they impose on everyone else, which makes honest reporting optimal regardless of what others do. The catch is that VCG normally requires searching over all possible blocks, which is impossible here. Their fix is to require the block-building algorithm to commit to a fixed, bid-independent set of candidate blocks and to be able to find the best one and each "what if this bundle were absent" counterfactual quickly. Any algorithm with those two properties yields a truthful, efficient mechanism — that is the theoretical backbone.
For the default algorithm, they first measured how Ethereum transactions actually conflict over storage. Finding that conflicts come in small clusters, they group transactions so that different groups don't interfere, then exhaustively search within each small group, with shortcuts for structures like sandwich bundles. They then tested this on historical transaction data.
Why This Matters
Impact on research. The paper reframes builder centralization as a mechanism-design failure rather than an inevitability of MEV, and shows how to extend VCG to a setting where the outcome space is exponential and the computation must finish in seconds. It also gives a formal vocabulary — searcher-DSIC, builder-DSIC, integration-resistance, welfare maximization, computational efficiency, individual rationality, budget balance — and is explicit that not all of these can hold simultaneously, which sets a concrete agenda for follow-up work.
Real-world applications:
- Ethereum block building and PBS: replacing or augmenting the MEV-Boost auction with a mechanism that treats searchers as first-class participants.
- TEE-based builder markets: the design parallels BuilderNet-style deployments, where algorithms and order flows run in cloud-hosted TEEs.
- MEV transparency and compliance: since payments and refunds are settled by transactions in the output block rather than opaque off-chain transfers, the flow of MEV becomes observable to the ecosystem.
- Apps and wallets that route order flow: users and searchers who currently rely on private deals would instead submit bids into a single governed pool with equal access.
Industry relevance. The paper speaks directly to builders, searchers, and proposers who currently operate in a market where two builders produce most blocks. Its claim that the design rewards algorithmic innovation over private partnerships is an argument aimed at operators deciding where to invest engineering effort, and its trust assumptions (perfect TEE, DoS out of scope, honest proposers) are stated plainly enough for practitioners to judge deployability.
Future Directions
- Quantifying welfare loss from the restricted block set. The paper's guarantee is truthfulness for any algorithm satisfying its two properties; how much welfare the 53.2%-optimal default algorithm leaves on the table, and how much builder-submitted algorithms recover, is the natural next measurement.
- Hardening against the
block.coinbasesignaling trick. The authors concede integration cannot be fully eliminated. Designing constructions that make state-dependent integration detection harder — beyond making the default algorithm indistinguishable from other algorithms — remains open. - Relaxing the perfect-TEE and honest-proposer assumptions. Side channels, physical attacks, and denial-of-service are explicitly out of scope, as are proposer deviations; each is a separate trust question for a production deployment.
- Reputation and admission control for algorithm submissions. The paper mentions a reputation system to curb DoS from algorithm submissions, analogous to existing searcher reputation mechanisms, but does not specify or evaluate one.
Target Audience
Researchers and protocol engineers working on blockchain security, MEV, and proposer–builder separation; mechanism-design theorists interested in computationally constrained VCG variants; and builders, searchers, or infrastructure operators who want to understand whether integration advantages can be designed away. Readers should be comfortable with auction theory and Ethereum execution semantics — this is not an introductory paper.
Authors’ abstract
Block space on the blockchain is scarce and must be allocated efficiently through block building. However, Ethereum's current block-building ecosystem, MEV-Boost, has become highly centralized due to integration, which distorts competition, reduces blockspace efficiency, and obscures MEV flow transparency. To guarantee equitability and economic efficiency in block building, we propose $\mathrm{Boost+}$, a system that decouples the process into collecting and ordering…
The full abstract on arXiv