The Base-Rate Effect on LLM Benchmark Performance: Disambiguating Test-Taking Strategies from Benchmark Performance
Kyle Moore, Jesse Roberts, Thao Pham, Oseremhen Ewaleifoh, Douglas Fisher
Abstract
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.
BibTeX
@inproceedings{moore-etal-2024-base,
title = "The Base-Rate Effect on {LLM} Benchmark Performance: Disambiguating Test-Taking Strategies from Benchmark Performance",
author = "Moore, Kyle and
Roberts, Jesse and
Pham, Thao and
Ewaleifoh, Oseremhen and
Fisher, Douglas",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2024",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-emnlp.126/",
doi = "10.18653/v1/2024.findings-emnlp.126",
pages = "2283--2288"
}