MathDial: Dialogue Tutoring Dataset Grounded in Math Word Problems

MathDial provides a challenging testbed for creating dialogue tutoring models that can reason over complex information and student confusions. The dataset is grounded in math word problems and is designed to address the gap where models solve problems accurately but fail at teaching. It was created by eth-nlped and the associated paper was published in 2023.

Use Cases

• Training dialogue agents for math tutoring based on pedagogical properties mentioned in the description
• Benchmarking model faithfulness and equity in educational AI using grounded math problems
• Studying how AI systems handle student confusions during instructional dialogue

Strengths

• Dataset is grounded in math word problems, providing a structured reasoning challenge
• Explicitly designed with rich pedagogical properties for tutoring research

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

huggingface, author eth-nlped

Collection Method

Likely constructed from math word problems and simulated or annotated tutoring dialogues.

Freshness

Last updated 2025-02-26 07:34:18; freshness should be verified