Saved in:
Bibliographic Details
Main Authors: Moore, Kyle, Roberts, Jesse, Pham, Thao, Ewaleifoh, Oseremhen, Fisher, Doug
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2406.11634
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913523044450304
author Moore, Kyle
Roberts, Jesse
Pham, Thao
Ewaleifoh, Oseremhen
Fisher, Doug
author_facet Moore, Kyle
Roberts, Jesse
Pham, Thao
Ewaleifoh, Oseremhen
Fisher, Doug
contents Cloze testing is a common method for measuring the behavior of large language models on a number of benchmark tasks. Using the MMLU dataset, we show that the base-rate probability (BRP) differences across answer tokens are significant and affect task performance ie. guess A if uncertain. We find that counterfactual prompting does sufficiently mitigate the BRP effect. The BRP effect is found to have a similar effect to test taking strategies employed by humans leading to the conflation of task performance and test-taking ability. We propose the Nvr-X-MMLU task, a variation of MMLU, which helps to disambiguate test-taking ability from task performance and reports the latter.
format Preprint
id arxiv_https___arxiv_org_abs_2406_11634
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle The Base-Rate Effect on LLM Benchmark Performance: Disambiguating Test-Taking Strategies from Benchmark Performance
Moore, Kyle
Roberts, Jesse
Pham, Thao
Ewaleifoh, Oseremhen
Fisher, Doug
Computation and Language
Artificial Intelligence
Cloze testing is a common method for measuring the behavior of large language models on a number of benchmark tasks. Using the MMLU dataset, we show that the base-rate probability (BRP) differences across answer tokens are significant and affect task performance ie. guess A if uncertain. We find that counterfactual prompting does sufficiently mitigate the BRP effect. The BRP effect is found to have a similar effect to test taking strategies employed by humans leading to the conflation of task performance and test-taking ability. We propose the Nvr-X-MMLU task, a variation of MMLU, which helps to disambiguate test-taking ability from task performance and reports the latter.
title The Base-Rate Effect on LLM Benchmark Performance: Disambiguating Test-Taking Strategies from Benchmark Performance
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2406.11634