Synthetic Eggs in Many Baskets: The Impact of Synthetic Data Diversity on LLM Fine-Tuning
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914514108153856 |
|---|---|
| author | Schaffelder, Max Gatt, Albert |
| author_facet | Schaffelder, Max Gatt, Albert |
| contents | As synthetic data becomes widely used in language model development, understanding its impact on model behavior is crucial. This paper investigates the impact of the diversity of sources of synthetic data on fine-tuned large language models. We focus on three key dimensions: distribution collapse, adversarial robustness, and self-preference bias. Our findings reveal that fine-tuning models on synthetic data from diverse sources can mitigate distribution collapse, preserving the breadth of the output distribution and the diversity of the output text. Furthermore, while both human and synthetic fine-tuning data can remove safeguards, we observe a tendency for higher output quality in the latter case, thus making outputs potentially more usable and dangerous. Finally, we also find evidence that fine-tuning reduces self-preference bias, with human data being the most effective, followed by multi-source synthetic data. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_01490 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Synthetic Eggs in Many Baskets: The Impact of Synthetic Data Diversity on LLM Fine-Tuning Schaffelder, Max Gatt, Albert Computation and Language As synthetic data becomes widely used in language model development, understanding its impact on model behavior is crucial. This paper investigates the impact of the diversity of sources of synthetic data on fine-tuned large language models. We focus on three key dimensions: distribution collapse, adversarial robustness, and self-preference bias. Our findings reveal that fine-tuning models on synthetic data from diverse sources can mitigate distribution collapse, preserving the breadth of the output distribution and the diversity of the output text. Furthermore, while both human and synthetic fine-tuning data can remove safeguards, we observe a tendency for higher output quality in the latter case, thus making outputs potentially more usable and dangerous. Finally, we also find evidence that fine-tuning reduces self-preference bias, with human data being the most effective, followed by multi-source synthetic data. |
| title | Synthetic Eggs in Many Baskets: The Impact of Synthetic Data Diversity on LLM Fine-Tuning |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2511.01490 |