Research map / Large language models / Agents, code and math
From Solver to Tutor: Evaluating the Pedagogical Intelligence of LLMs with KMP-Bench
Weikang Shi, Houxing Ren, Junting Pan, Aojun Zhou, Ke Wang, Zimu Lu, Yunqiao Yang, Yuxuan Hu, Linda Wei, Mingjie Zhan, Hongsheng Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| fine-tune on KMP-Pile tutoring dialogues | Qwen2.5-Math-7B-Inst · KMP-Dialogue | Overall Acc: 23.6 -> 37.0 | Improved |
| fine-tune on KMP-Pile tutoring dialogues | Qwen2.5-Math-7B-Inst · KMP-Dialogue | Overall Judgement: 21.0 -> 40.4 | Improved |
| fine-tune on KMP-Pile instead of MathChat_sync data | Qwen2.5-Math-7B · KMP-Dialogue | Overall Acc: 15.3 -> 37.0 | Improved |
| fine-tune on KMP-Pile tutoring dialogues | Qwen2.5-Math-7B-Inst · KMP-EC | MR-Score: 3.4 -> 49.0 | Improved |
| fine-tune on KMP-Pile tutoring dialogues | Qwen2.5-Math-7B-Inst · KMP-Gen Follow-up | pass rate: 48.6 -> 81.1 | Improved |
| fine-tune on KMP-Pile tutoring dialogues | Qwen2.5-Math-7B-Inst · KMP-Gen Similar | pass rate: 52.8 -> 83.2 | Improved |
| fine-tune on KMP-Pile tutoring dialogues | Qwen2.5-Math-7B-Inst · KMP-FQA | accuracy: 76.1 -> 77.5 | Improved |
| use general-purpose model instead of math-tuned model | Qwen2.5-72B · KMP-Dialogue | Overall Acc: 30.8 -> 58.3 | Improved |
| use math-tuned model instead of general-purpose model | Qwen2.5-72B · KMP-Gen Follow-up | pass rate: 81.3 -> 86.3 | Mixed |
| Gemini-2.0-Flash evaluator vs manual grading (ablation) | KMP-Dialogue evaluation · 300-instance subset | Alignment Rate: 89.8% average alignment | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SocraticLM — SocraticLM: Exploring Socratic Personalized Teaching with Large Language Models (NeurIPS 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Leveraging Online Olympiad-Level Math Problems for LLMs Training and Contamination-Resistant Evaluation (ICML 2025 · similarity 0.89)
- Advancing Mathematical Reasoning in Language Models: The Impact of Problem-Solving Data, Data Synthesis Methods, and Training Stages (ICLR 2025 · similarity 0.89)
- A Careful Examination of Large Language Model Performance on Grade School Arithmetic (NeurIPS 2024 · similarity 0.88)
- CogMath: Assessing LLMs' Authentic Mathematical Ability from a Human Cognitive Perspective (ICML 2025 · similarity 0.88)
- Language Models as Science Tutors (ICML 2024 · similarity 0.88)
- UGMathBench: A Diverse and Dynamic Benchmark for Undergraduate-Level Mathematical Reasoning with Large Language Models (ICLR 2025 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
