Can Out-of-Distribution Evaluations Uncover Reliance on Shortcuts? A Case Study in Question Answering

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216224%3A14330%2F25%3A00142633" target="_blank" >RIV/00216224:14330/25:00142633 - isvavai.cz</a>

Result on the web

<a href="https://aclanthology.org/2025.findings-emnlp.1232/" target="_blank" >https://aclanthology.org/2025.findings-emnlp.1232/</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.18653/v1/2025.findings-emnlp.1232" target="_blank" >10.18653/v1/2025.findings-emnlp.1232</a>

Result language

angličtina

Original language name

Can Out-of-Distribution Evaluations Uncover Reliance on Shortcuts? A Case Study in Question Answering

Original language description

A majority of recent work in AI assesses models' generalization capabilities through the lens of performance on out-of-distribution (OOD) datasets. Despite their practicality, such evaluations build upon a strong assumption: that OOD evaluations can capture and reflect upon possible failures in a real-world deployment. In this work, we challenge this assumption and confront the results obtained from OOD evaluations with a set of specific failure modes documented in existing question-answering (QA) models, referred to as a reliance on spurious features or prediction shortcuts. We find that different datasets used for OOD evaluations in QA provide an estimate of models' robustness to shortcuts that have a vastly different quality, some largely under-performing even a simple, in-distribution evaluation. We partially attribute this to the observation that spurious shortcuts are shared across ID+OOD datasets, but also find cases where a dataset's quality for training and evaluation is largely disconnected. Our work underlines limitations of commonly-used OOD-based evaluations of generalization, and provides methodology and recommendations for evaluating generalization within and beyond QA more robustly.

Czech name

—

Czech description

—

Type

D - Article in proceedings

CEP classification

—

OECD FORD branch

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Project

—

Continuities

S - Specificky vyzkum na vysokych skolach

Publication year

2025

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection

Findings of the Association for Computational Linguistics: EMNLP 2025

ISBN

9798891763357

ISSN

—

e-ISSN

—

Number of pages

8

Pages from-to

22628-22635

Publisher name

Association for Computational Linguistics

Place of publication

Suzhou, China

Event location

Suzhou, China

Event date

Jan 1, 2025

Type of event by nationality

WRD - Celosvětová akce

UT code for WoS article

—