SocraticMATH: 6,846 Chinese Primary School Math Tutoring Dialogues

6,846 multi-turn Socratic tutoring conversations covering 513 primary school math knowledge points. This dataset was created by Ding et al. for the paper 'Boosting Large Language Models with Socratic Method for Conversational Mathematics Teaching' (CIKM '24). It is hosted on Hugging Face by the user ulises-c.

Use Cases

• Training conversational AI tutors based on Socratic teaching dialogues.
• Evaluating large language models on multi-turn, pedagogical reasoning tasks.
• Analyzing the structure of educational dialogues for primary school mathematics.
• Developing fine-tuning datasets for culturally and linguistically specific educational AI.

Strengths

• Contains 6,846 multi-turn conversations, providing substantial volume for training.
• Covers 513 distinct primary school math knowledge points, offering topic breadth.
• Created for a peer-reviewed CIKM 2024 publication, suggesting academic rigor.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the specific variant is unknown, which may limit suitability assessment.
• Data may reflect geographic and educational bias inherent to the Chinese primary school context.

Provenance

Source

Paper 'Boosting Large Language Models with Socratic Method for Conversational Mathematics Teaching' (CIKM '24) by Ding et al.

Collection Method

Likely constructed for research on Socratic math tutoring.

Freshness

Last updated 2026-05-27 01:32:54; freshness should be verified.

Geography

Chinese primary school context.