Synthetic Pre-Pre-Training Improves Language Model Robustness to Noisy Pre-Training Data
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866916000154255360 |
|---|---|
| author | Guo, Xu Peng, Runyu Tong, Jian Zhou, Yunhua Lv, Haijun Lu, Zhihui Guo, Qipeng |
| author_facet | Guo, Xu Peng, Runyu Tong, Jian Zhou, Yunhua Lv, Haijun Lu, Zhihui Guo, Qipeng |
| contents | Large language models (LLMs) rely on web-scale corpora for pre-training. The noise inherent in these datasets tends to obscure meaningful patterns and ultimately degrade model performance. Data curation mitigates but cannot eliminate such noise, so pre-training corpora remain noisy in practice. We therefore study whether a lightweight pre-pre-training (PPT) stage based on synthetic data with learnable temporal structure helps resist noisy data during the pre-training (PT) stage. Across various corruption settings, our method consistently improves robustness to noise during PT, with larger relative gains at higher noise levels. For a 1B-parameter model, a synthetic PPT stage with only 65M tokens achieves the same final loss as the baseline while using up to 49\% fewer natural-text PT tokens across different noise levels. Mechanistic analyses suggest PPT does not immediately suppress attention to noisy tokens. Rather, PPT-initialized models gradually downweight attention between corrupted tokens during noisy PT. This indicates that synthetic PPT inhibits noise self-modeling and shapes the subsequent optimization trajectory. Code is available at https://github.com/guox18/formal-language-prepretraining. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_10129 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Synthetic Pre-Pre-Training Improves Language Model Robustness to Noisy Pre-Training Data Guo, Xu Peng, Runyu Tong, Jian Zhou, Yunhua Lv, Haijun Lu, Zhihui Guo, Qipeng Computation and Language Large language models (LLMs) rely on web-scale corpora for pre-training. The noise inherent in these datasets tends to obscure meaningful patterns and ultimately degrade model performance. Data curation mitigates but cannot eliminate such noise, so pre-training corpora remain noisy in practice. We therefore study whether a lightweight pre-pre-training (PPT) stage based on synthetic data with learnable temporal structure helps resist noisy data during the pre-training (PT) stage. Across various corruption settings, our method consistently improves robustness to noise during PT, with larger relative gains at higher noise levels. For a 1B-parameter model, a synthetic PPT stage with only 65M tokens achieves the same final loss as the baseline while using up to 49\% fewer natural-text PT tokens across different noise levels. Mechanistic analyses suggest PPT does not immediately suppress attention to noisy tokens. Rather, PPT-initialized models gradually downweight attention between corrupted tokens during noisy PT. This indicates that synthetic PPT inhibits noise self-modeling and shapes the subsequent optimization trajectory. Code is available at https://github.com/guox18/formal-language-prepretraining. |
| title | Synthetic Pre-Pre-Training Improves Language Model Robustness to Noisy Pre-Training Data |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2605.10129 |