EduResearchBench - Theory and Survey: leaderboard
Metric: Judge score (1-5; 5 atomic tasks, Chinese only, mean of Doubao-seed-1.6 and DeepSeek-V3.2 critique-then-score ratings). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 | 3.63 |
| 2 | Gemini 3 Flash | 3.21 |
| 3 | Qwen 2.5 72B Instruct | 2.42 |
| 4 | LongWriter-glm4-9B | 1.77 |
| 5 | Llama 3 70B Instruct | 1.5 |
Interactive version: theaggregate.ai/benchmark?slug=eduresearchbench-theory-and-survey · How It Works · Data refreshed daily, snapshot 2026-09-25.