Saved in:
Bibliographic Details
Main Authors: Healey, Jennifer, Byrum, Laurie, Akhtar, Md Nadeem, Bhargava, Surabhi, Sinha, Moumita
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.03053
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910929200873472
author Healey, Jennifer
Byrum, Laurie
Akhtar, Md Nadeem
Bhargava, Surabhi
Sinha, Moumita
author_facet Healey, Jennifer
Byrum, Laurie
Akhtar, Md Nadeem
Bhargava, Surabhi
Sinha, Moumita
contents LLM evaluation is challenging even the case of base models. In real world deployments, evaluation is further complicated by the interplay of task specific prompts and experiential context. At scale, bias evaluation is often based on short context, fixed choice benchmarks that can be rapidly evaluated, however, these can lose validity when the LLMs' deployed context differs. Large scale human evaluation is often seen as too intractable and costly. Here we present our journey towards developing a semi-automated bias evaluation framework for free text responses that has human insights at its core. We discuss how we developed an operational definition of bias that helped us automate our pipeline and a methodology for classifying bias beyond multiple choice. We additionally comment on how human evaluation helped us uncover problematic templates in a bias benchmark.
format Preprint
id arxiv_https___arxiv_org_abs_2505_03053
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Developing A Framework to Support Human Evaluation of Bias in Generated Free Response Text
Healey, Jennifer
Byrum, Laurie
Akhtar, Md Nadeem
Bhargava, Surabhi
Sinha, Moumita
Computation and Language
Artificial Intelligence
LLM evaluation is challenging even the case of base models. In real world deployments, evaluation is further complicated by the interplay of task specific prompts and experiential context. At scale, bias evaluation is often based on short context, fixed choice benchmarks that can be rapidly evaluated, however, these can lose validity when the LLMs' deployed context differs. Large scale human evaluation is often seen as too intractable and costly. Here we present our journey towards developing a semi-automated bias evaluation framework for free text responses that has human insights at its core. We discuss how we developed an operational definition of bias that helped us automate our pipeline and a methodology for classifying bias beyond multiple choice. We additionally comment on how human evaluation helped us uncover problematic templates in a bias benchmark.
title Developing A Framework to Support Human Evaluation of Bias in Generated Free Response Text
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2505.03053