SurveyReview - Readability: leaderboard

Metric: Mean squared error of predicted readability scores (0 to 16, lower is better; zero-shot LLM-as-a-judge with the unified dimension-specific rubric prompt on the full survey text, against scores annotated from real peer-review reports of the test-split surveys (MOPRD, F1000Research, OpenReview) on the {-2, -1, +1, +2} scale). Source: arxiv.org. Saturation forecast: Around December 2026. 6 models tracked.

Top models

#ModelScore
1GPT-5.22.13
2Claude Opus 4.5 (Thinking)2.91
3Qwen 3 32B3.05
4GLM-4.73.43
5Gemini 3 Pro (Preview)3.84
6DeepSeek V3.24.78

Interactive version: theaggregate.ai/benchmark?slug=surveyreview-readability · How It Works · Data refreshed daily, snapshot 2026-09-26.