GPT-5 mini · LongFact-Objects(GPT-5 developer release; no tools) 1.8% · 附件证据(ev-gpt5-mini-card-thinking-browse-off-longfact-objects)

GPT-5 mini 的 LongFact-Objects · GPT-5 developer release; no tools 官方出处。GPT-5 thinking · browsing off · 1.8%。见《GPT-5 System Card (current 60-page revision)》Current PDF p.14, Figure 3; model/benchmark bar label visually checked.。证据 ev-gpt5-mini-card-thinking-browse-off-longfact-objects。

原文与定位

GPT-5 mini · LongFact-Objects(GPT-5 developer release; no tools),GPT-5 thinking · browsing off配置,Hallucination rate 1.8%(具体成绩公布时间未确认)。见《GPT-5 System Card (current 60-page revision)》Current PDF p.14, Figure 3; model/benchmark bar label visually checked.。 查看成绩条件

Average Hallucination Rate (Browsing Disabled): gpt-5-thinking-mini; LongFact-Objects: 1.8%.
官方材料
GPT-5 System Card (current 60-page revision) · PDF 第 14 页
定位
附件 · Current PDF p.14, Figure 3; model/benchmark bar label visually checked.
材料版本
Current canonical PDF dated August 13, 2025; text pp.1-60 fully read; all 45 table/figure pages visually inspected, plus p.31; detailed scope in 2026-11-16-review-gpt-early/PROGRESS.md. Exact current revision date unknown.
版本公布
未确认
披露时间
未确认 · 首发/后补未确认
采集时间
历史依据
暂无可靠历史版本依据
引用文本

GPT-5 mini · LongFact-Objects(GPT-5 developer release; no tools),GPT-5 thinking · browsing off配置,Hallucination rate 1.8%(具体成绩公布时间未确认)。官方出处 GPT-5 System Card (current 60-page revision)(附件,Current canonical PDF dated August 13, 2025; text pp.1-60 fully read; all 45 table/figure pages visually inspected, plus p.31; detailed scope in 2026-11-16-review-gpt-early/PROGRESS.md. Exact current revision date unknown.)Current PDF p.14, Figure 3; model/benchmark bar label visually checked.,https://cdn.openai.com/gpt-5-system-card.pdf#page=14。Benchlog https://benchlog.lekshome.cc/evidence/gpt-5-mini/ev-gpt5-mini-card-thinking-browse-off-longfact-objects/

关联成绩与条件

GPT-5 mini · LongFact-Objects 成绩与配置
指标与成绩配置与评测条件证据关系
Hallucination rate1.8%原文 1.8% · 单位 %
公布 未确认
GPT-5 thinking · browsing off
设置
System Card thinking model; browsing disabled; OpenAI o3 claim extraction and web-based factuality grading.
未知项
图未注明具体reasoning effort、采样参数或重复次数。
工具
工具条件未记录
推理预算
未记录
框架
未记录
协议
LongFact-Objects; thinking model; browsing off; claim-level hallucination rate; OpenAI o3 grader.
条件
Explicit Figure 2/3 numeric labels; effort not inferred from a different API announcement run; no curve estimation.

直接支持该成绩

在排名中查看此成绩