Research map / Large language models / Alignment and preferences
Teach2Eval: An Interaction-Driven LLMs Evaluation Method via Teaching Effectiveness
Yuhang Zhou, Xutian Chen, Yixin Cao, Yuchen Ni, Yu He, Siyu Tian, Xiang Liu, Yunwen Chen, Guangnan Ye, Xipeng Qiu, Hongfeng Chai
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 |
|---|---|---|---|
| evaluate via teaching weaker students instead of direct answer scoring | LLM evaluation · Chatbot Arena | Spearman correlation: 0.734 -> 0.944 | Improved |
| evaluate via teaching weaker students instead of direct answer scoring | LLM evaluation · LiveBench | Spearman correlation: 0.861 -> 0.975 | Improved |
| evaluate via teaching instead of direct scoring | LLM evaluation · Chatbot Arena | Kendall's Tau: 0.558 -> 0.853 | Improved |
| evaluate via teaching instead of direct scoring | LLM evaluation · LiveBench | Kendall's Tau: 0.695 -> 0.886 | Improved |
| remove one of four student models at random (ablation) | Teach2Eval · Chatbot Arena | Spearman correlation: 0.926-0.936 | No clear effect |
| increase cumulative sample size, Teach2Eval vs direct (ablation) | Teach2Eval · LiveBench | Kendall's Tau: 0.83 vs near 0.64 | Improved |
| more guidance rounds (up to six) (ablation) | student models · guidance rounds | student gain | No clear effect |
| teacher reveals the answer vs not (ablation) | Teach2Eval ranking · answer revelation subsets | Spearman vs Teach2Eval: 0.947 -> 0.921 | No clear effect |
| distill GPT-4o answers into models (contamination-like fine-tuning) | six LLMs · AA vs CA | Application and Comprehensive Ability | Mixed |
Closest papers by meaning
- EvaLearn: Quantifying the Learning Capability and Efficiency of LLMs via Sequential Problem Solving (NeurIPS 2025 · similarity 0.91)
- DistiLLM-2: A Contrastive Approach Boosts the Distillation of LLMs (ICML 2025 · similarity 0.90)
- Evaluating Large Language Models at Evaluating Instruction Following (ICLR 2024 · similarity 0.90)
- RocketEval: Efficient automated LLM evaluation via grading checklist (ICLR 2025 · similarity 0.90)
- MINT: Evaluating LLMs in Multi-turn Interaction with Tools and Language Feedback (ICLR 2024 · similarity 0.90)
- Re-evaluating Open-ended Evaluation of Large Language Models (ICLR 2025 · similarity 0.89)
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.
