# Still Not There: Can LLMs Outperform Smaller Task-Specific Seq2Seq Models on the Poetry-to-Prose Conversion Task?

arXiv: https://arxiv.org/abs/2511.08145 · HTML: https://mlsanity.com/paper/2511.08145/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 15 of its 32 references that are in the map.
- First posted on arXiv 11 Nov 2025.

## Builds on (most cited first, 15 of 15)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Large Language Models are Zero-Shot Reasoners](https://mlsanity.com/paper/2205.11916/) (2022)
- [A Call for Clarity in Reporting BLEU Scores](https://mlsanity.com/paper/1804.08771/) (2018)
- [ByT5: Towards a token-free future with pre-trained byte-to-byte models](https://mlsanity.com/paper/2105.13626/) (2021)
- [Few-shot Learning with Multilingual Language Models](https://mlsanity.com/paper/2112.10668/) (2021)
- [One Model is All You Need: ByT5-Sanskrit, a Unified Model for Sanskrit NLP Tasks](https://mlsanity.com/paper/2409.13920/) (2024)
- [TransLIST: A Transformer-Based Linguistically Informed Sanskrit Tokenizer](https://mlsanity.com/paper/2210.11753/) (2022)
- [A Little Pretraining Goes a Long Way: A Case Study on Dependency Parsing Task for Low-resource Morphologically Rich Languages](https://mlsanity.com/paper/2102.06551/) (2021)
- [SanskritShala: A Neural Sanskrit NLP Toolkit with Web-Based Interface for Pedagogical and Annotation Purposes](https://mlsanity.com/paper/2302.09527/) (2023)
- [Chandomitra: Towards Generating Structured Sanskrit Poetry from Natural Language Inputs](https://mlsanity.com/paper/2506.00815/) (2025)
- [Neural Approaches for Data Driven Dependency Parsing in Sanskrit](https://mlsanity.com/paper/2004.08076/) (2020)
- [A Novel Multi-Task Learning Approach for Context-Sensitive Compound Type Identification in Sanskrit](https://mlsanity.com/paper/2208.10310/) (2022)
- [Systematic Investigation of Strategies Tailored for Low-Resource Settings for Low-Resource Dependency Parsing](https://mlsanity.com/paper/2201.11374/) (2022)
- [Vedavani: A Benchmark Corpus for ASR on Vedic Sanskrit Poetry](https://mlsanity.com/paper/2506.00145/) (2025)
- [DepNeCTI: Dependency-based Nested Compound Type Identification for Sanskrit](https://mlsanity.com/paper/2310.09501/) (2023)

## Built on by (most cited first, 2 of 2)

- [Translating Classical Poetry into Modern Prose](https://mlsanity.com/paper/2606.02806/) (2026)
- [Padamitra: Grounded Glossary Generation for Classical Sanskrit](https://mlsanity.com/paper/2608.25038/) (2026)

## Abstract (arXiv)

Large Language Models (LLMs) are increasingly treated as universal, general-purpose solutions across NLP tasks, particularly in English. But does this assumption hold for low-resource, morphologically rich languages such as Sanskrit? We address this question by comparing instruction-tuned and in-context-prompted LLMs with smaller task-specific encoder-decoder models on the Sanskrit poetry-to-prose conversion task. This task is intrinsically challenging: Sanskrit verse exhibits free word order combined with rigid metrical constraints, and its conversion to canonical prose (anvaya) requires multi-step reasoning involving compound segmentation, dependency resolution, and syntactic linearisation. This makes it an ideal testbed to evaluate whether LLMs can surpass specialised models. For LLMs, we apply instruction fine-tuning on general-purpose models and design in-context learning templates grounded in Paninian grammar and classical commentary heuristics. For task-specific modelling, we fully fine-tune a ByT5-Sanskrit Seq2Seq model. Our experiments show that domain-specific fine-tuning of ByT5-Sanskrit significantly outperforms all instruction-driven LLM approaches. Human evaluation strongly corroborates this result, with scores exhibiting high correlation with Kendall's Tau scores. Additionally, our prompting strategies provide an alternative to fine-tuning when domain-specific verse corpora are unavailable, and the task-specific Seq2Seq model demonstrates robust generalisation on out-of-domain evaluations.