Gemini 4 Argon · SimpleQA Verified(Kaggle official leaderboard) 78.8% · 附件证据(ev-g4a-simpleqa-f1)

Gemini 4 Argon 的 SimpleQA Verified · Kaggle official leaderboard 官方出处。Gemini 4 Argon report results; effort unspecified · 78.8%。见《Gemini 4 Argon Safety & Security Report》PDF p.192 (printed p.191), Figure 5.12, SimpleQA Verified F1 Score (%) bar label。证据 ev-g4a-simpleqa-f1。

原文与定位

Gemini 4 Argon · SimpleQA Verified(Kaggle official leaderboard),Gemini 4 Argon report results; effort unspecified配置,F1 score 78.8%(具体成绩公布时间未确认)。见《Gemini 4 Argon Safety & Security Report》PDF p.192 (printed p.191), Figure 5.12, SimpleQA Verified F1 Score (%) bar label。 查看成绩条件

SimpleQA Verified (F1 score, %) | Gemini 4 Argon 78.8.
官方材料
Gemini 4 Argon Safety & Security Report · PDF 第 192 页
定位
附件 · PDF p.192 (printed p.191), Figure 5.12, SimpleQA Verified F1 Score (%) bar label
材料版本
Official 217-page PDF; cover dated October 2026; all pp.1-217 read; conventional-score pages PDF pp.178, 192-193 rendered and visually checked
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

Gemini 4 Argon · SimpleQA Verified(Kaggle official leaderboard),Gemini 4 Argon report results; effort unspecified配置,F1 score 78.8%(具体成绩公布时间未确认)。官方出处 Gemini 4 Argon Safety & Security Report(附件,Official 217-page PDF; cover dated October 2026; all pp.1-217 read; conventional-score pages PDF pp.178, 192-193 rendered and visually checked)PDF p.192 (printed p.191), Figure 5.12, SimpleQA Verified F1 Score (%) bar label,https://storage.googleapis.com/deepmind-media/gemini/gemini_4_safety_report.pdf#page=192。Benchlog https://benchlog.lekshome.cc/evidence/gemini-4-argon/ev-g4a-simpleqa-f1/

关联成绩与条件

Gemini 4 Argon · SimpleQA Verified 成绩与配置
指标与成绩配置与评测条件证据关系
F1 score78.8%原文 78.8% · 单位 %
公布 未确认
Gemini 4 Argon report results; effort unspecified
设置
Gemini 4 Argon; reasoning effort, sampling, repeat count, and pass policy are not stated in the Safety & Security Report. Search grounding is recorded only where the report specifies it.
未知项
安全报告未注明这些常规评测的思考档位、采样参数、重复次数或 pass 聚合方式。
工具
工具条件未记录
推理预算
未记录
框架
未记录
协议
1,000 human-verified, de-duplicated short-answer questions; closed-book; responses graded correct, incorrect, or abstained; F1 is the harmonic mean of overall and attempted-question accuracy.
条件
Parametric-knowledge setting; reasoning effort, sampling, run count, and pass policy are not stated.

直接支持该成绩