How to Synthesize Text Data without Model Collapse?
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866913862723305472 |
|---|---|
| author | Zhu, Xuekai Cheng, Daixuan Li, Hengli Zhang, Kaiyan Hua, Ermo Lv, Xingtai Ding, Ning Lin, Zhouhan Zheng, Zilong Zhou, Bowen |
| author_facet | Zhu, Xuekai Cheng, Daixuan Li, Hengli Zhang, Kaiyan Hua, Ermo Lv, Xingtai Ding, Ning Lin, Zhouhan Zheng, Zilong Zhou, Bowen |
| contents | Model collapse in synthetic data indicates that iterative training on self-generated data leads to a gradual decline in performance. With the proliferation of AI models, synthetic data will fundamentally reshape the web data ecosystem. Future GPT-$\{n\}$ models will inevitably be trained on a blend of synthetic and human-produced data. In this paper, we focus on two questions: what is the impact of synthetic data on language model training, and how to synthesize data without model collapse? We first pre-train language models across different proportions of synthetic data, revealing a negative correlation between the proportion of synthetic data and model performance. We further conduct statistical analysis on synthetic data to uncover distributional shift phenomenon and over-concentration of n-gram features. Inspired by the above findings, we propose token editing on human-produced data to obtain semi-synthetic data. As a proof of concept, we theoretically demonstrate that token-level editing can prevent model collapse, as the test error is constrained by a finite upper bound. We conduct extensive experiments on pre-training from scratch, continual pre-training, and supervised fine-tuning. The results validate our theoretical proof that token-level editing improves model performance. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_14689 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | How to Synthesize Text Data without Model Collapse? Zhu, Xuekai Cheng, Daixuan Li, Hengli Zhang, Kaiyan Hua, Ermo Lv, Xingtai Ding, Ning Lin, Zhouhan Zheng, Zilong Zhou, Bowen Computation and Language Artificial Intelligence Machine Learning Model collapse in synthetic data indicates that iterative training on self-generated data leads to a gradual decline in performance. With the proliferation of AI models, synthetic data will fundamentally reshape the web data ecosystem. Future GPT-$\{n\}$ models will inevitably be trained on a blend of synthetic and human-produced data. In this paper, we focus on two questions: what is the impact of synthetic data on language model training, and how to synthesize data without model collapse? We first pre-train language models across different proportions of synthetic data, revealing a negative correlation between the proportion of synthetic data and model performance. We further conduct statistical analysis on synthetic data to uncover distributional shift phenomenon and over-concentration of n-gram features. Inspired by the above findings, we propose token editing on human-produced data to obtain semi-synthetic data. As a proof of concept, we theoretically demonstrate that token-level editing can prevent model collapse, as the test error is constrained by a finite upper bound. We conduct extensive experiments on pre-training from scratch, continual pre-training, and supervised fine-tuning. The results validate our theoretical proof that token-level editing improves model performance. |
| title | How to Synthesize Text Data without Model Collapse? |
| topic | Computation and Language Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2412.14689 |