Rethinking Data: Towards Better Performing Domain-Specific Small Language Models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866912259817603072 |
|---|---|
| author | Nazarov, Boris Frolova, Darya Lubarsky, Yackov Gaissinski, Alexei Kisilev, Pavel |
| author_facet | Nazarov, Boris Frolova, Darya Lubarsky, Yackov Gaissinski, Alexei Kisilev, Pavel |
| contents | Fine-tuning of Large Language Models (LLMs) for downstream tasks, performed on domain-specific data has shown significant promise. However, commercial use of such LLMs is limited by the high computational cost required for their deployment at scale. On the other hand, small Language Models (LMs) are much more cost effective but have subpar performance in a similar setup. This paper presents our approach to finetuning a small LM, that reaches high accuracy in multiple choice question answering task. We achieve this by improving data quality at each stage of the LM training pipeline. In particular, we start with data structuring resulting in extraction of compact, semantically meaningful text chunks used by a retriever. This allows more efficient knowledge digestion by the LM. Further, we improve the retrieved context by training a lightweight Chunk Re-Ranker (CRR) that generates more accurate relative relevance chunk scores. Finally, we improve the model generalization ability by merging the models fine-tuned with different parameters on different data subsets. We present detailed procedure descriptions, and corresponding experimental findings that show the improvements of each one of the proposed techniques. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2503_01464 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Rethinking Data: Towards Better Performing Domain-Specific Small Language Models Nazarov, Boris Frolova, Darya Lubarsky, Yackov Gaissinski, Alexei Kisilev, Pavel Computation and Language Artificial Intelligence Fine-tuning of Large Language Models (LLMs) for downstream tasks, performed on domain-specific data has shown significant promise. However, commercial use of such LLMs is limited by the high computational cost required for their deployment at scale. On the other hand, small Language Models (LMs) are much more cost effective but have subpar performance in a similar setup. This paper presents our approach to finetuning a small LM, that reaches high accuracy in multiple choice question answering task. We achieve this by improving data quality at each stage of the LM training pipeline. In particular, we start with data structuring resulting in extraction of compact, semantically meaningful text chunks used by a retriever. This allows more efficient knowledge digestion by the LM. Further, we improve the retrieved context by training a lightweight Chunk Re-Ranker (CRR) that generates more accurate relative relevance chunk scores. Finally, we improve the model generalization ability by merging the models fine-tuned with different parameters on different data subsets. We present detailed procedure descriptions, and corresponding experimental findings that show the improvements of each one of the proposed techniques. |
| title | Rethinking Data: Towards Better Performing Domain-Specific Small Language Models |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2503.01464 |