Mining Hidden Thoughts from Texts: Evaluating Continual Pretraining with Synthetic Data for LLM Reasoning
Fuente:
arXiv
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866915288043225088 |
|---|---|
| author | Ishibashi, Yoichi Yano, Taro Oyamada, Masafumi |
| author_facet | Ishibashi, Yoichi Yano, Taro Oyamada, Masafumi |
| contents | Large Language Models (LLMs) have demonstrated significant improvements in reasoning capabilities through supervised fine-tuning and reinforcement learning. However, when training reasoning models, these approaches are primarily applicable to specific domains such as mathematics and programming, which imposes fundamental constraints on the breadth and scalability of training data. In contrast, continual pretraining (CPT) offers the advantage of not requiring task-specific signals. Nevertheless, how to effectively synthesize training data for reasoning and how such data affect a wide range of domains remain largely unexplored. This study provides a detailed evaluation of Reasoning CPT, a form of CPT that uses synthetic data to reconstruct the hidden thought processes underlying texts, based on the premise that texts are the result of the author's thinking process. Specifically, we apply Reasoning CPT to Gemma2-9B using synthetic data with hidden thoughts derived from STEM and Law corpora, and compare it to standard CPT on the MMLU benchmark. Our analysis reveals that Reasoning CPT consistently improves performance across all evaluated domains. Notably, reasoning skills acquired in one domain transfer effectively to others; the performance gap with conventional methods widens as problem difficulty increases, with gains of up to 8 points on the most challenging problems. Furthermore, models trained with hidden thoughts learn to adjust the depth of their reasoning according to problem difficulty. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_10182 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Mining Hidden Thoughts from Texts: Evaluating Continual Pretraining with Synthetic Data for LLM Reasoning Ishibashi, Yoichi Yano, Taro Oyamada, Masafumi Computation and Language Machine Learning Large Language Models (LLMs) have demonstrated significant improvements in reasoning capabilities through supervised fine-tuning and reinforcement learning. However, when training reasoning models, these approaches are primarily applicable to specific domains such as mathematics and programming, which imposes fundamental constraints on the breadth and scalability of training data. In contrast, continual pretraining (CPT) offers the advantage of not requiring task-specific signals. Nevertheless, how to effectively synthesize training data for reasoning and how such data affect a wide range of domains remain largely unexplored. This study provides a detailed evaluation of Reasoning CPT, a form of CPT that uses synthetic data to reconstruct the hidden thought processes underlying texts, based on the premise that texts are the result of the author's thinking process. Specifically, we apply Reasoning CPT to Gemma2-9B using synthetic data with hidden thoughts derived from STEM and Law corpora, and compare it to standard CPT on the MMLU benchmark. Our analysis reveals that Reasoning CPT consistently improves performance across all evaluated domains. Notably, reasoning skills acquired in one domain transfer effectively to others; the performance gap with conventional methods widens as problem difficulty increases, with gains of up to 8 points on the most challenging problems. Furthermore, models trained with hidden thoughts learn to adjust the depth of their reasoning according to problem difficulty. |
| title | Mining Hidden Thoughts from Texts: Evaluating Continual Pretraining with Synthetic Data for LLM Reasoning |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2505.10182 |