GPT-6.2 Astra · User-flagged factuality · Refreshed 1,000 de-identified conversations: 229 retained cases and 771 new cases; separate general factuality and reported-issue graders; not comparable with earlier system-card versions · 官方披露
GPT-6.2 Astra 的 User-flagged factuality · Refreshed 1,000 de-identified conversations: 229 retained cases and 771 new cases; separate general factuality and reported-issue graders; not comparable with earlier system-card versions 记录。本次发布已收录官方披露,暂无成绩记录,附 1 条来源定位。
成绩与评测条件
已收录官方披露,暂无成绩记录。
出处与披露
- HTML §3.3.1 Performance in Cases Flagged by Users / Figure 25; PDF physical pages 72–73 / printed pages 71–72 / §3.3.0.1. 图中没有精确点值标签,HTML仅嵌入静态PNG,不按坐标估值。 附件 · GPT-6.2 Astra and Sol System Card
测试版本与指标
别名 Performance in Cases Flagged by Users
- Responses with at least one major factual error · % · 越低越好
- Responses reproducing the user-reported error · % · 越低越好