Research map / Large language models / Agents, code and math
EssayBench: Evaluating Large Language Models in Multi-Genre Chinese Essay Writing
Fan Gao, Dongyuan Li, Ding Xia, Fei Mi, Yasheng Wang, Lifeng Shang, Baojun 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 dependency-based weights (ablation) | EssayBench scoring, DeepSeek-V3 judge · Human agreement (overall) | Spearman rho: 0.646 -> 0.667 | Improved |
| add dependency-based weights (ablation) | EssayBench scoring, GPT-4o judge · Human agreement (overall) | Spearman rho: 0.706 -> 0.733 | Improved |
| add dependency-based weights (ablation) | EssayBench scoring, DeepSeek-R1 judge · Human agreement (overall) | Spearman rho: 0.803 -> 0.816 | Improved |
| add dependency-based weights (ablation) | EssayBench scoring, DeepSeek-V3 judge · Human agreement (overall) | Kendall tau: 0.529 -> 0.549 | Improved |
| replace Align-Score with fine-grained genre-oriented protocol | EssayBench scoring, DeepSeek-V3 judge · Human agreement (overall) | Spearman rho: 0.674 -> 0.667 | Hurt |
| replace Align-Score with fine-grained genre-oriented protocol | EssayBench scoring, GPT-4o judge · Human agreement (overall) | Spearman rho: 0.628 -> 0.733 | Improved |
| replace Align-Score with fine-grained genre-oriented protocol | EssayBench scoring, DeepSeek-R1 judge · Human agreement (overall) | Spearman rho: 0.749 -> 0.816 | Improved |
| use a stronger judge (DeepSeek-R1 instead of DeepSeek-V3) | EssayBench scoring · Human agreement (overall) | Spearman rho: 0.667 -> 0.816 | Improved |
| replace Align-Score with proposed protocol | DeepSeek-R1 judge · Sensitivity high&medium | Mann-Whitney U proportion: 0.64 -> 0.79 | Improved |
| replace Align-Score with proposed protocol | DeepSeek-V3 judge · Sensitivity medium&low | Mann-Whitney U proportion: 0.90 -> 0.78 | Hurt |
| replace Align-Score with proposed protocol | DeepSeek-R1 judge · Sensitivity high&low | Mean difference: 2.41 -> 3.49 | Improved |
| upgrade Claude-3.5-sonnet to Claude-3.7-sonnet | Claude · EssayBench overall | score: 75.4 -> 76.6 | Improved |
Closest papers by meaning
- WenMind: A Comprehensive Benchmark for Evaluating Large Language Models in Chinese Classical Literature and Language Arts (NeurIPS 2024 · similarity 0.87)
- WritingBench: A Comprehensive Benchmark for Generative Writing (NeurIPS 2025 · similarity 0.85)
- CK12: A Rounded K12 Knowledge Graph Based Benchmark for Chinese Holistic Cognition Evaluation (AAAI 2024 · similarity 0.85)
- Does Writing with Language Models Reduce Content Diversity? (ICLR 2024 · similarity 0.84)
- KAES: Multi-aspect Shared Knowledge Finding and Aligning for Cross-prompt Automated Scoring of Essay Traits (AAAI 2025 · similarity 0.84)
- Unveiling the Tapestry of Automated Essay Scoring: A Comprehensive Investigation of Accuracy, Fairness, and Generalizability (AAAI 2024 · similarity 0.84)
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.
