Unmasking the Factual-Conceptual Gap in Persian Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sakhaeirad, Alireza, Ma'manpoosh, Ali, Hemmat, Arshia
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908841368616960
author Sakhaeirad, Alireza
Ma'manpoosh, Ali
Hemmat, Arshia
author_facet Sakhaeirad, Alireza
Ma'manpoosh, Ali
Hemmat, Arshia
contents While emerging Persian NLP benchmarks have expanded into pragmatics and politeness, they rarely distinguish between memorized cultural facts and the ability to reason about implicit social norms. We introduce DivanBench, a diagnostic benchmark focused on superstitions and customs, arbitrary, context-dependent rules that resist simple logical deduction. Through 315 questions across three task types (factual retrieval, paired scenario verification, and situational reasoning), we evaluate seven Persian LLMs and reveal three critical failures: most models exhibit severe acquiescence bias, correctly identifying appropriate behaviors but failing to reject clear violations; continuous Persian pretraining amplifies this bias rather than improving reasoning, often degrading the model's ability to discern contradictions; and all models show a 21\% performance gap between retrieving factual knowledge and applying it in scenarios. These findings demonstrate that cultural competence requires more than scaling monolingual data, as current models learn to mimic cultural patterns without internalizing the underlying schemas.
format Preprint
id arxiv_https___arxiv_org_abs_2602_17623
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Unmasking the Factual-Conceptual Gap in Persian Language Models
Sakhaeirad, Alireza
Ma'manpoosh, Ali
Hemmat, Arshia
Computation and Language
While emerging Persian NLP benchmarks have expanded into pragmatics and politeness, they rarely distinguish between memorized cultural facts and the ability to reason about implicit social norms. We introduce DivanBench, a diagnostic benchmark focused on superstitions and customs, arbitrary, context-dependent rules that resist simple logical deduction. Through 315 questions across three task types (factual retrieval, paired scenario verification, and situational reasoning), we evaluate seven Persian LLMs and reveal three critical failures: most models exhibit severe acquiescence bias, correctly identifying appropriate behaviors but failing to reject clear violations; continuous Persian pretraining amplifies this bias rather than improving reasoning, often degrading the model's ability to discern contradictions; and all models show a 21\% performance gap between retrieving factual knowledge and applying it in scenarios. These findings demonstrate that cultural competence requires more than scaling monolingual data, as current models learn to mimic cultural patterns without internalizing the underlying schemas.
title Unmasking the Factual-Conceptual Gap in Persian Language Models
topic Computation and Language
url https://arxiv.org/abs/2602.17623