Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2409.12599 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910612313866240 |
|---|---|
| author | Pieper, Tom Ballout, Mohamad Krumnack, Ulf Heidemann, Gunther Kühnberger, Kai-Uwe |
| author_facet | Pieper, Tom Ballout, Mohamad Krumnack, Ulf Heidemann, Gunther Kühnberger, Kai-Uwe |
| contents | This paper explores the enhancement of small language models through strategic dataset augmentation via ChatGPT-3.5-Turbo, in the domain of Natural Language Inference (NLI). By employing knowledge distillation-based techniques and synthetic dataset augmentation, we aim to bridge the performance gap between large language models (LLMs) and small language models (SLMs) without the immense cost of human annotation. Our methods involve two forms of rationale generation--information extraction and informed reasoning--to enrich the ANLI dataset. We then fine-tune T5-Small on these augmented datasets, evaluating its performance against an established benchmark. Our findings reveal that the incorporation of synthetic rationales significantly improves the model's ability to comprehend natural language, leading to 1.3\% and 2.3\% higher classification accuracy, respectively, on the ANLI dataset, demonstrating the potential of leveraging LLMs for dataset augmentation. This approach not only enhances the performance of smaller models on complex tasks but also introduces a cost-effective method for fine-tuning smaller language models. By advancing our understanding of knowledge distillation and fine-tuning strategies, this work contributes to the ongoing effort to create more capable and efficient NLP systems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2409_12599 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Enhancing SLM via ChatGPT and Dataset Augmentation Pieper, Tom Ballout, Mohamad Krumnack, Ulf Heidemann, Gunther Kühnberger, Kai-Uwe Computation and Language This paper explores the enhancement of small language models through strategic dataset augmentation via ChatGPT-3.5-Turbo, in the domain of Natural Language Inference (NLI). By employing knowledge distillation-based techniques and synthetic dataset augmentation, we aim to bridge the performance gap between large language models (LLMs) and small language models (SLMs) without the immense cost of human annotation. Our methods involve two forms of rationale generation--information extraction and informed reasoning--to enrich the ANLI dataset. We then fine-tune T5-Small on these augmented datasets, evaluating its performance against an established benchmark. Our findings reveal that the incorporation of synthetic rationales significantly improves the model's ability to comprehend natural language, leading to 1.3\% and 2.3\% higher classification accuracy, respectively, on the ANLI dataset, demonstrating the potential of leveraging LLMs for dataset augmentation. This approach not only enhances the performance of smaller models on complex tasks but also introduces a cost-effective method for fine-tuning smaller language models. By advancing our understanding of knowledge distillation and fine-tuning strategies, this work contributes to the ongoing effort to create more capable and efficient NLP systems. |
| title | Enhancing SLM via ChatGPT and Dataset Augmentation |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2409.12599 |