Papers by Pralay Manna
A Case Study of Cross-Lingual Zero-Shot Generalization for Classical Languages in LLMs (2025.findings-acl)
Copied to clipboard
V.S.D.S.Mahesh Akavarapu, Hrishikesh Terdalkar, Pramit Bhattacharyya, Shubhangi Agarwal, Dr. Vishakha Deulgaonkar, Chaitali Dangarikar, Pralay Manna, Arnab Bhattacharya
| Challenge: | Large Language Models (LLMs) have demonstrated remarkable generalization capabilities across diverse tasks and languages. |
| Approach: | They focus on named entity recognition and machine translation into English to examine factors affecting cross-lingual zero-shot generalization. |
| Outcome: | The proposed models perform better than fine-tuned baselines on out-of-domain data, but smaller models struggle with niche or abstract entity types. |