When Can Voting Help, Hurt, or Change Course? Exact Structure of Binary Test-Time Aggregation

Yi Liu · 2026

Read When Can Voting Help, Hurt, or Change Course? Exact Structure of Binary Test-Time Aggregation by Yi Liu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-05592.txt.

Description

Asking a model the same question many times and taking the majority answer often improves accuracy. Yi Liu examines exactly when that trick helps, hurts, or flips direction as votes accumulate. Classical theory, assuming a fixed skill level per question, implies a tidy monotone story: above a threshold, more votes always help.

Liu shows the real picture is richer. Under a latent distribution of per-question success probabilities, voting curves can bend repeatedly, even infinitely often in an explicit construction. The full distribution determines the curve, but the curve cannot be run backward to recover the distribution.

Catalog record

CREATORYi Liu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE59 KB
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2605.05592

Browse the full library · About the Tower of Babel