On Benchmark Hacking in ML Contests: Modeling, Insights and Design

TL;DR

This paper analyzes benchmark hacking in ML contests, modeling contestant strategies, equilibrium behavior, and how reward structures influence efforts, supported by empirical evidence.

Contribution

It introduces a game-theoretic framework for benchmark hacking, characterizes equilibrium effort allocations, and explores how reward skewness affects contest outcomes.

Findings

Low-type contestants always engage in benchmark hacking.

More skewed rewards incentivize higher effort from top contestants.

Empirical evidence supports the theoretical predictions.

Abstract

Benchmark hacking refers to tuning a machine learning model to score highly on certain evaluation criteria without improving true generalization or faithfully solving the intended problem. We study this phenomenon in a generic machine learning contest, where each contestant chooses two types of effort: creative effort that improves model capability as desired by the contest host, and mechanistic effort that only improves the model's fitness to the particular task in contest without contributing to true generalization. We establish the existence of a symmetric monotone pure strategy equilibrium in this competition game. It also provides a natural definition of benchmark hacking in this strategic context by comparing a player's equilibrium effort allocation to that of a single-agent baseline scenario. Under our definition, contestants with types below certain threshold (low types) always engage in benchmark hacking, whereas those above the threshold do not. Furthermore, we show that more skewed reward structures (favoring top-ranked contestants) can elicit more desirable contest outcomes. We also provide empirical evidence to support our theoretical predictions.