GPT-5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations) · 附件证据(ev-gpt52-healthbench)

GPT-5.2 的 HealthBench · 48,000+ rubric items across 5,000 multi-turn conversations 官方出处。见《GPT-5.2 System Card Update PDF | OpenAI》PDF p. 9 (printed p. 8), Table 5 / HealthBench row。证据 ev-gpt52-healthbench。

原文与定位

GPT-5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations),GPT-5.2 Instant (system card)配置,Length-adjusted score 47.6066%(成绩公布 2025-12-11)。见《GPT-5.2 System Card Update PDF | OpenAI》PDF p. 9 (printed p. 8), Table 5 / HealthBench row。 查看成绩条件

GPT-5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations),GPT-5.2 Thinking (system card)配置,Length-adjusted score 63.3379%(成绩公布 2025-12-11)。见《GPT-5.2 System Card Update PDF | OpenAI》PDF p. 9 (printed p. 8), Table 5 / HealthBench row。 查看成绩条件

Table 5: HealthBench. HealthBench 0.482146 | 0.476066 | 0.639872 | 0.633379; HealthBench Hard 0.208091 | 0.171893 | 0.404925 | 0.420389; HealthBench Consensus 0.949287 | 0.943521 | 0.959684 | 0.945020. Columns: gpt-5.1-instant | gpt-5.2-instant | gpt-5.1-thinking | gpt-5.2-thinking.
官方材料
GPT-5.2 System Card Update PDF | OpenAI · PDF 第 9 页
定位
附件 · PDF p. 9 (printed p. 8), Table 5 / HealthBench row
材料版本
官方 27 页 PDF,完整核读 PDF p. 1-27,含图表页均已实际渲染检查;封面标注 December 11, 2025。
版本公布
披露时间
· 已有首发依据
采集时间
历史依据
暂无可靠历史版本依据
引用文本

GPT-5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations),GPT-5.2 Instant (system card)配置,Length-adjusted score 47.6066%(成绩公布 2025-12-11);GPT-5.2 · HealthBench(48,000+ rubric items across 5,000 multi-turn conversations),GPT-5.2 Thinking (system card)配置,Length-adjusted score 63.3379%(成绩公布 2025-12-11)。官方出处 GPT-5.2 System Card Update PDF | OpenAI(附件,官方 27 页 PDF,完整核读 PDF p. 1-27,含图表页均已实际渲染检查;封面标注 December 11, 2025。)PDF p. 9 (printed p. 8), Table 5 / HealthBench row,https://cdn.openai.com/pdf/3a4153c8-c748-4b71-8e31-aecbde944f8d/oai_5_2_system-card.pdf#page=9。Benchlog https://benchlog.lekshome.cc/evidence/gpt-5-2/ev-gpt52-healthbench/

关联成绩与条件

GPT-5.2 · HealthBench 成绩与配置
指标与成绩配置与评测条件证据关系
Length-adjusted score47.6066%原文 0.476066 · 单位 %
公布
GPT-5.2 Instant (system card)
设置
System-card HealthBench table names GPT-5.2 Instant; serving configuration is not stated.
未知项
Exact reasoning effort and serving configuration are not reported.
工具
工具条件未记录
推理预算
未记录
框架
未记录
协议
HealthBench, 5,000 potentially multi-turn conversations scored with example-specific rubrics; source decimal fraction converted to percentage.
条件
GPT-5.2 Instant; reasoning effort and serving configuration are not stated.

直接支持该成绩

Length-adjusted score63.3379%原文 0.633379 · 单位 %
公布
GPT-5.2 Thinking (system card)
设置
System-card HealthBench table names GPT-5.2 Thinking; serving configuration is not stated.
未知项
Exact reasoning effort and serving configuration are not reported.
工具
工具条件未记录
推理预算
未记录
框架
未记录
协议
HealthBench, 5,000 potentially multi-turn conversations scored with example-specific rubrics; source decimal fraction converted to percentage.
条件
GPT-5.2 Thinking; reasoning effort and serving configuration are not stated.

直接支持该成绩