Welfare, Improvability, and Variance: A Principal-Agent Approach to Optimal Benchmark Item Aggregation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Haupt, Andreas, Hartenstein, Justin, Reuel, Anka, Kochenderfer, Mykel, Koyejo, Sanmi
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910271793004544
author Haupt, Andreas
Hartenstein, Justin
Reuel, Anka
Kochenderfer, Mykel
Koyejo, Sanmi
author_facet Haupt, Andreas
Hartenstein, Justin
Reuel, Anka
Kochenderfer, Mykel
Koyejo, Sanmi
contents AI benchmarks have well-documented limitations, with prior work examining contamination, saturation, and construct underspecification. Aggregation has received far less attention: benchmarks are typically summarized by uniformly averaging item-level scores, implicitly treating every test item as equally valuable. We model benchmarking as a multitask principal-agent game and show that the welfare loss from a benchmark is determined jointly by three item-level primitives: alignment with normative welfare priorities, marginal improvability, and performance variance. We translate the theory into an audit framework that ranks items along each of these three axes, and apply it to OLMES items using WORKBank for welfare, the EvoLM 4B suite for improvability, and the PolyPythias 410M panel for variance. The framework surfaces items that are Pareto-inferior within OLMES subject to a pro-worker welfare operationalization. All code is available at https://github.com/stair-lab/principal-agent-benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2605_30916
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Welfare, Improvability, and Variance: A Principal-Agent Approach to Optimal Benchmark Item Aggregation
Haupt, Andreas
Hartenstein, Justin
Reuel, Anka
Kochenderfer, Mykel
Koyejo, Sanmi
Machine Learning
Computer Science and Game Theory
Theoretical Economics
AI benchmarks have well-documented limitations, with prior work examining contamination, saturation, and construct underspecification. Aggregation has received far less attention: benchmarks are typically summarized by uniformly averaging item-level scores, implicitly treating every test item as equally valuable. We model benchmarking as a multitask principal-agent game and show that the welfare loss from a benchmark is determined jointly by three item-level primitives: alignment with normative welfare priorities, marginal improvability, and performance variance. We translate the theory into an audit framework that ranks items along each of these three axes, and apply it to OLMES items using WORKBank for welfare, the EvoLM 4B suite for improvability, and the PolyPythias 410M panel for variance. The framework surfaces items that are Pareto-inferior within OLMES subject to a pro-worker welfare operationalization. All code is available at https://github.com/stair-lab/principal-agent-benchmarks.
title Welfare, Improvability, and Variance: A Principal-Agent Approach to Optimal Benchmark Item Aggregation
topic Machine Learning
Computer Science and Game Theory
Theoretical Economics
url https://arxiv.org/abs/2605.30916