RBCorr: Response Bias Correction in Language Models

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Bhatt, Om, Ivanova, Anna A.
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866917272176558080
author Bhatt, Om
Ivanova, Anna A.
author_facet Bhatt, Om
Ivanova, Anna A.
contents Language models (LMs) are known to be prone to response biases, which present as option preference biases in fixed-response questions. It is therefore imperative to develop low-cost and effective response bias correction methods to improve LM performance and enable more accurate evaluations of model abilities. Here, we propose a simple response bias correction strategy ($\texttt{RBCorr}$) and test it on 12 open-weight language models using yes-no, entailment, and multiple choice questions. We show that response bias is prevalent in LMs pre-correction and that $\texttt{RBCorr}$ effectively eliminates bias and boosts model performance. We also explore the generalizability of bias behavior across models, datasets, and prompt formats, showing that LogProbs-based correction is highly dependent on all three of these aspects. Overall, $\texttt{RBCorr}$ is an easy-to-use method that can boost the performance of smaller LMs and ensure that LM performance on closed-response benchmarks aligns more closely with their true capabilities.
format Preprint
id arxiv_https___arxiv_org_abs_2602_12445
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle RBCorr: Response Bias Correction in Language Models
Bhatt, Om
Ivanova, Anna A.
Computation and Language
Machine Learning
I.2.7
Language models (LMs) are known to be prone to response biases, which present as option preference biases in fixed-response questions. It is therefore imperative to develop low-cost and effective response bias correction methods to improve LM performance and enable more accurate evaluations of model abilities. Here, we propose a simple response bias correction strategy ($\texttt{RBCorr}$) and test it on 12 open-weight language models using yes-no, entailment, and multiple choice questions. We show that response bias is prevalent in LMs pre-correction and that $\texttt{RBCorr}$ effectively eliminates bias and boosts model performance. We also explore the generalizability of bias behavior across models, datasets, and prompt formats, showing that LogProbs-based correction is highly dependent on all three of these aspects. Overall, $\texttt{RBCorr}$ is an easy-to-use method that can boost the performance of smaller LMs and ensure that LM performance on closed-response benchmarks aligns more closely with their true capabilities.
title RBCorr: Response Bias Correction in Language Models
topic Computation and Language
Machine Learning
I.2.7
url https://arxiv.org/abs/2602.12445