Socratic Students: Teaching Language Models to Learn by Asking Questions
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_ | 1866915709481648128 |
|---|---|
| author | Ambati, Rajeev Bhatt Niu, Tianyi Singh, Aashu Mishra, Shlok Chaturvedi, Snigdha Srivastava, Shashank |
| author_facet | Ambati, Rajeev Bhatt Niu, Tianyi Singh, Aashu Mishra, Shlok Chaturvedi, Snigdha Srivastava, Shashank |
| contents | Large language Models (LLMs) are usually used to answer questions, but many high-stakes applications (e.g., tutoring, clinical support) require the complementary skill of asking questions: detecting missing information, requesting clarifications, and using them to solve tasks. We study this skill in reasoning-heavy domains where progress depends on inquiry rather than factual recall. We define an interactive protocol where a student model engages a stronger teacher under a small turn budget. After each teacher reply, we evaluate the student on the original task with Pass@k. We propose Outcome-Driven Question optimization Strategy (ODQS ), a training framework that learns a questioning policy from downstream task outcomes. At each turn, we sample multiple candidate questions; query the teacher with each, then score the student's resulting performance. Using these scores, we train the student via supervised fine-tuning followed by Direct Preference Optimization (DPO), without any human labels. On GSM8K, HumanEval, and OpenCoder, ODQS produces large gains over interactive baselines, boosting Pass@5 by up to 54.7% (absolute) on math and 22.9% (absolute) on coding, and matching baseline performance in three fewer turns. Thus, question asking can be explicitly trained from task outcomes, improving both accuracy and efficiency in interactive reasoning. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_13102 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Socratic Students: Teaching Language Models to Learn by Asking Questions Ambati, Rajeev Bhatt Niu, Tianyi Singh, Aashu Mishra, Shlok Chaturvedi, Snigdha Srivastava, Shashank Artificial Intelligence Large language Models (LLMs) are usually used to answer questions, but many high-stakes applications (e.g., tutoring, clinical support) require the complementary skill of asking questions: detecting missing information, requesting clarifications, and using them to solve tasks. We study this skill in reasoning-heavy domains where progress depends on inquiry rather than factual recall. We define an interactive protocol where a student model engages a stronger teacher under a small turn budget. After each teacher reply, we evaluate the student on the original task with Pass@k. We propose Outcome-Driven Question optimization Strategy (ODQS ), a training framework that learns a questioning policy from downstream task outcomes. At each turn, we sample multiple candidate questions; query the teacher with each, then score the student's resulting performance. Using these scores, we train the student via supervised fine-tuning followed by Direct Preference Optimization (DPO), without any human labels. On GSM8K, HumanEval, and OpenCoder, ODQS produces large gains over interactive baselines, boosting Pass@5 by up to 54.7% (absolute) on math and 22.9% (absolute) on coding, and matching baseline performance in three fewer turns. Thus, question asking can be explicitly trained from task outcomes, improving both accuracy and efficiency in interactive reasoning. |
| title | Socratic Students: Teaching Language Models to Learn by Asking Questions |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2512.13102 |