Research map / Large language models / Language, safety and interpretability
A Compress-Expand Framework for Automatic Lesson Plan Generation
Shuyan Huang, Ying Zheng, Xiaoli Zeng, Zitao Liu
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 textbook expander (ablation) | CE-LessonPlan · Lesson plan generation | ROUGE-1: 11.606 -> 14.849 | Improved |
| add lesson plan compressor (ablation) | CE-LessonPlan · Lesson plan generation | ROUGE-1: 18.143 -> 14.849 | Hurt |
| add lesson plan compressor (ablation) | CE-LessonPlan · Lesson plan generation | Human evaluation average score: 2.273 -> 2.276 | No clear effect |
| add compressor and expander to the SFT+RAG generator (ablation) | CE-LessonPlan · Lesson plan generation | ROUGE-1: 7.334 -> 14.849 | Improved |
| add three retrieved lesson plans as few-shot examples | GPT-4o · Lesson plan generation | ROUGE-1: 6.708 -> 8.779 | Improved |
| add textbook information to the prompt | GPT-4o · Lesson plan generation | ROUGE-1: 6.708 -> 14.016 | Improved |
| add 3-shot retrieved lesson plans to textbook-enhanced prompt | GPT-4o · Lesson plan generation | ROUGE-1 and ROUGE-L: 14.016 -> 14.324 ROUGE-1; 12.775 -> 12.662 ROUGE-L | Mixed |
| raise ensemble coefficient alpha from 0.5 up to 0.8, then to 0.9 (ablation) | CE-LessonPlan · Lesson plan generation | ROUGE-1 and ROUGE-L | Mixed |
| use SFT lesson plan model instead of Qwen2.5-72B as generator (ablation) | CE-LessonPlan · Lesson plan generation | ROUGE-1: 14.511 -> 14.849 | Improved |
| use SFT lesson plan model instead of Qwen2.5-72B as compressor and expander (ablation) | CE-LessonPlan · Lesson plan generation | ROUGE-1: 11.964 -> 14.849 | Improved |
| use compress-expand framework instead of Self-Critique | lesson plan generator · Lesson plan generation | ROUGE-1: 13.317 -> 14.849 | Improved |
| use compress-expand framework instead of Self-Critique | lesson plan generator · Lesson plan generation | Human MAT score: 2.605 -> 1.933 | Hurt |
Closest papers by meaning
- MiniPLM: Knowledge Distillation for Pre-training Language Models (ICLR 2025 · similarity 0.84)
- Exploring Iterative Enhancement for Improving Learnersourced Multiple-Choice Question Explanations with Large Language Models (AAAI 2025 · similarity 0.83)
- DistiLLM-2: A Contrastive Approach Boosts the Distillation of LLMs (ICML 2025 · similarity 0.83)
- LongGenBench: Benchmarking Long-Form Generation in Long Context LLMs (ICLR 2025 · similarity 0.82)
- Speculative Knowledge Distillation: Bridging the Teacher-Student Gap Through Interleaved Sampling (ICLR 2025 · similarity 0.82)
- How Teachers Can Use Large Language Models and Bloom’s Taxonomy to Create Educational Quizzes (AAAI 2024 · similarity 0.82)
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.
