2026-11-05 –, Lightning Stage
Ordering transactions creates value, and today that value is split by position in the pipeline rather than by contribution: builders get squeezed, users get nothing, and rebate splits are set by convention. We argue this split isn't a matter of taste — four ordinary fairness axioms force a single rule. We show how to compute it for proposers, builders, and order flow providers in milliseconds, and where paying users back turns provably hard.
Rebate splits in MEV today (90/10, pro-rata) are chosen by convention, and the recurring objection to a principled alternative is that "fair" splits are exponential to compute. This talk refutes that. Taking contribution-based fairness seriously — the Shapley value, singled out by four standard axioms — we show the split is exactly computable in closed form for proposers, builders, and order flow providers, fast enough for a 12-second slot. On the user side we show the problem is provably hard (sub-exponential distinct marginal contributions) and give a sampling algorithm with PAC guarantees. The finding that matters now: as exclusive order flow concentrates power among a few block builders, paying contribution rather than position is both fair and computationally practical — not just aspirational.
Dr. Sujit Gujar is an Associate Professor and CA Technologies Faculty Chair at IIIT Hyderabad, leading the Machine Learning Lab. His research develops game-theoretic foundations for modern AI and distributed AI/blockchains, ensuring fairness in these systems, spanning 125+ peer-reviewed publications and 15+ US patents. He earned his Ph.D. from the Indian Institute of Science, Bangalore, winning the IISc Alumni Medal for best CSA thesis, and previously researched at EPFL and Xerox Research.