AI写作拟真度基准The Slop Index正式上线
Start here
2026年11月7日，基准评测平台 The Slop Index 正式上线，旨在评估各款前沿大语言模型写作风格与真实人类表达的差距，评选哪款模型最没有“AI味”。该基准共评估了58款前沿大模型。在评测机制上，评分完全不依赖大语言模型裁判，而是将生成内容与2022年之前的人类职场邮件、推文和文章进行对比；综合评分体系由人类盲测投票（占40%）以及简洁度、模板化、韵律感和典型AI句式（各占15%）构成，最终输出0到100分的偏离得分。
AI-generated from reports · Updated 45 min ago
Report timeline
Follow the reports to see different sides of the story.
- Hacker News · AISelectedThe Slop Index Launches: Benchmarking Which AI Models Sound the Least Like AI
The Slop Index, a new evaluation benchmark, has officially launched to measure the gap between the writing styles of 58 frontier large language models and authentic human expression using quantitative metrics and blind human evaluations. Completely avoiding LLM-as-a-judge methodologies, it compares model-generated text against pre-2022 human workplace emails, tweets, and articles. The benchmark calculates a divergence score from 0 to 100 based on blind human votes (40%), along with conciseness, formulaic phrasing, cadence, and typical AI sentence structures (15% each).
Story heat trend
Not enough continuous data yet to draw a trend.