语言模型生成彩票号码的重复集中现象研究
原标题：Ask a Language Model for Lottery Numbers: Concentration in Repeated Six-of-49 Outputs
AI 摘要
研究者在 arXiv 上发布论文，用 1-49 选 6 的彩票号码请求测试六种语言模型配置（gpt-5.6-luna、claude-sonnet-5、gemini-3.7-flash、grok-4.5、mistral-large、Perplexity sonar），共 1200 次调用中 1184 次返回有效号码。结果显示模型输出高度集中：有效多样性仅 9.9-18.0，远低于均匀采样模拟的 46.6-46.7 下限，且重复整张彩票的概率远高于独立均匀采样。论文强调这仅证明在测试部署设置下存在集中现象，未识别其机制，也未验证其他提示、温度或工具配置下的表现。
正文节选
Ask a Language Model for Lottery Numbers: Concentration in Repeated Six-of-49 Outputs Abstract We evaluate six language-model configurations on requests for six distinct random integers from 1-49. Across 1,200 attempted calls using four English prompt variants, 1,184 responses yielded valid tickets. Effective diversity of number frequencies ranged from 9.9 to 18.0, compared with simulated fifth-percentile thresholds of 46.6-46.7 under independent uniform six-of-49 sampling at the corresponding s