Research map / Large language models / Agents, code and math
From Diagnosis to Generalization: A Cognitive Approach to Data Selection for Educational LLMs
Yuxiang Guo, Yan Zhuang, Qi Liu, Zhenya Huang, Xianquan Wang, Liyang He, Jiatong Li, Rui Li, Shijin Wang
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 |
|---|---|---|---|
| add behavioral probing before MIRT estimation (ablation) | CASS · EduData (LLaMA3-8B) | accuracy: out-of-domain 0.5531 -> 0.5779 | Improved |
| use Fisher Information selection instead of highest-difficulty selection (ablation) | CASS · EduData (LLaMA3-8B) | accuracy: out-of-domain 0.5562 -> 0.5779 | Improved |
| use MIRT diagnosis instead of lowest-accuracy heuristic (ablation) | CASS · EduData (LLaMA3-8B) | accuracy: out-of-domain 0.5494 -> 0.5779 | Improved |
| add easy-to-hard curriculum fine-tuning (ablation) | CASS · EduData (LLaMA3-8B) | accuracy: out-of-domain 0.5601 -> 0.5779 | Improved |
| select 10% data with CASS vs random selection | CASS · EduData (LLaMA3-8B, Table 1) | accuracy: out-of-domain 0.5489 -> 0.5779 | Improved |
| CASS vs strongest baseline LESS | data selection · EduData (LLaMA3-8B) | accuracy: out-of-domain 0.5580 -> 0.5779 | Improved |
| CASS vs LESS on Qwen2.5-7B | data selection · EduData (Qwen2.5-7B) | accuracy: out-of-domain 0.8733 -> 0.8864 | Improved |
| DSIR selection vs base model | LLaMA3-8B · EduData (LLaMA3-8B) | accuracy: out-of-domain 0.5286 -> 0.5307 | No clear effect |
| random selection of 10% data | LLaMA3-8B · EduData (LLaMA3-8B) | improvement over base: in-domain negative in History and Chemistry | Mixed |
Closest papers by meaning
- Knowledge Is Power: Harnessing Large Language Models for Enhanced Cognitive Diagnosis (AAAI 2025 · similarity 0.88)
- Unveiling the Learning Mind of Language Models: A Cognitive Framework and Empirical Study (NeurIPS 2025 · similarity 0.88)
- DistiLLM-2: A Contrastive Approach Boosts the Distillation of LLMs (ICML 2025 · similarity 0.87)
- MASS: Mathematical Data Selection via Skill Graphs for Pretraining Large Language Models (ICML 2025 · similarity 0.87)
- Generating Computational Cognitive models using Large Language Models (NeurIPS 2025 · similarity 0.87)
- What Do Learning Dynamics Reveal About Generalization in LLM Mathematical Reasoning? (ICML 2025 · similarity 0.87)
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.
