Saved in:
Bibliographic Details
Main Authors: Zgreabăn, Mădălina, Deoskar, Tejaswini, Abzianidze, Lasha
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2510.24295
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912673966325760
author Zgreabăn, Mădălina
Deoskar, Tejaswini
Abzianidze, Lasha
author_facet Zgreabăn, Mădălina
Deoskar, Tejaswini
Abzianidze, Lasha
contents In recent years, many generalization benchmarks have shown language models' lack of robustness in natural language inference (NLI). However, manually creating new benchmarks is costly, while automatically generating high-quality ones, even by modifying existing benchmarks, is extremely difficult. In this paper, we propose a methodology for automatically generating high-quality variants of original NLI problems by replacing open-class words, while crucially preserving their underlying reasoning. We dub our generalization test as MERGE (Minimal Expression-Replacements GEneralization), which evaluates the correctness of models' predictions across reasoning-preserving variants of the original problem. Our results show that NLI models' perform 4-20% worse on variants, suggesting low generalizability even on such minimally altered problems. We also analyse how word class of the replacements, word probability, and plausibility influence NLI models' performance.
format Preprint
id arxiv_https___arxiv_org_abs_2510_24295
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MERGE: Minimal Expression-Replacement GEneralization Test for Natural Language Inference
Zgreabăn, Mădălina
Deoskar, Tejaswini
Abzianidze, Lasha
Computation and Language
In recent years, many generalization benchmarks have shown language models' lack of robustness in natural language inference (NLI). However, manually creating new benchmarks is costly, while automatically generating high-quality ones, even by modifying existing benchmarks, is extremely difficult. In this paper, we propose a methodology for automatically generating high-quality variants of original NLI problems by replacing open-class words, while crucially preserving their underlying reasoning. We dub our generalization test as MERGE (Minimal Expression-Replacements GEneralization), which evaluates the correctness of models' predictions across reasoning-preserving variants of the original problem. Our results show that NLI models' perform 4-20% worse on variants, suggesting low generalizability even on such minimally altered problems. We also analyse how word class of the replacements, word probability, and plausibility influence NLI models' performance.
title MERGE: Minimal Expression-Replacement GEneralization Test for Natural Language Inference
topic Computation and Language
url https://arxiv.org/abs/2510.24295