A Case Study of Cross-Lingual Zero-Shot Generalization for Classical Languages in LLMs

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Akavarapu, V. S. D. S. Mahesh, Terdalkar, Hrishikesh, Bhattacharyya, Pramit, Agarwal, Shubhangi, Deulgaonkar, Vishakha, Manna, Pralay, Dangarikar, Chaitali, Bhattacharya, Arnab
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866910977343094784
author Akavarapu, V. S. D. S. Mahesh
Terdalkar, Hrishikesh
Bhattacharyya, Pramit
Agarwal, Shubhangi
Deulgaonkar, Vishakha
Manna, Pralay
Dangarikar, Chaitali
Bhattacharya, Arnab
author_facet Akavarapu, V. S. D. S. Mahesh
Terdalkar, Hrishikesh
Bhattacharyya, Pramit
Agarwal, Shubhangi
Deulgaonkar, Vishakha
Manna, Pralay
Dangarikar, Chaitali
Bhattacharya, Arnab
contents Large Language Models (LLMs) have demonstrated remarkable generalization capabilities across diverse tasks and languages. In this study, we focus on natural language understanding in three classical languages -- Sanskrit, Ancient Greek and Latin -- to investigate the factors affecting cross-lingual zero-shot generalization. First, we explore named entity recognition and machine translation into English. While LLMs perform equal to or better than fine-tuned baselines on out-of-domain data, smaller models often struggle, especially with niche or abstract entity types. In addition, we concentrate on Sanskrit by presenting a factoid question-answering (QA) dataset and show that incorporating context via retrieval-augmented generation approach significantly boosts performance. In contrast, we observe pronounced performance drops for smaller LLMs across these QA tasks. These results suggest model scale as an important factor influencing cross-lingual generalization. Assuming that models used such as GPT-4o and Llama-3.1 are not instruction fine-tuned on classical languages, our findings provide insights into how LLMs may generalize on these languages and their consequent utility in classical studies.
format Preprint
id arxiv_https___arxiv_org_abs_2505_13173
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Case Study of Cross-Lingual Zero-Shot Generalization for Classical Languages in LLMs
Akavarapu, V. S. D. S. Mahesh
Terdalkar, Hrishikesh
Bhattacharyya, Pramit
Agarwal, Shubhangi
Deulgaonkar, Vishakha
Manna, Pralay
Dangarikar, Chaitali
Bhattacharya, Arnab
Computation and Language
I.2.7
Large Language Models (LLMs) have demonstrated remarkable generalization capabilities across diverse tasks and languages. In this study, we focus on natural language understanding in three classical languages -- Sanskrit, Ancient Greek and Latin -- to investigate the factors affecting cross-lingual zero-shot generalization. First, we explore named entity recognition and machine translation into English. While LLMs perform equal to or better than fine-tuned baselines on out-of-domain data, smaller models often struggle, especially with niche or abstract entity types. In addition, we concentrate on Sanskrit by presenting a factoid question-answering (QA) dataset and show that incorporating context via retrieval-augmented generation approach significantly boosts performance. In contrast, we observe pronounced performance drops for smaller LLMs across these QA tasks. These results suggest model scale as an important factor influencing cross-lingual generalization. Assuming that models used such as GPT-4o and Llama-3.1 are not instruction fine-tuned on classical languages, our findings provide insights into how LLMs may generalize on these languages and their consequent utility in classical studies.
title A Case Study of Cross-Lingual Zero-Shot Generalization for Classical Languages in LLMs
topic Computation and Language
I.2.7
url https://arxiv.org/abs/2505.13173