AI2026/09/22@ScienceDesk
Google Publishes Attention Is All You Need
Googleplex, 1600, Amphitheatre Parkway, Mountain View, Santa Clara County, California, 94043, United StatesGoogle…
开始于 2017/06/12已确认可信度 88%9.3年前arXiv's^[1] own metadata for 1706.03762 gives citation_date 2017/06/12, and Wikipedia's^[2][3] article on the paper says the same: "On 2017-06-12, the original (100M-parameter) encoder-decoder transformer model was published in the 'Attention is all you need' paper."
Eight researchers at Google post a paper replacing recurrence and convolution with self-attention alone. The transformer architecture it introduces becomes the basis of nearly every large language model that follows.
0.67
AI2026/09/19@TechDesk
OpenAI Releases Open-Weight gpt-oss-120b and gpt-oss-20b
OpenAI
开始于 2025/08/05 已确认可信度 88%1.1年前OpenAI's^[1][2] own model release notes date this entry August 5, 2025
OpenAI releases two open-weight reasoning models, gpt-oss-120b and gpt-oss-20b, for teams that want to run and customize models themselves.
0.66
AI2026/09/19@TechDesk
OpenAI 发布 o3-mini
OpenAI
开始于 2025/01/31 已确认可信度 92%1.6年前OpenAI^[1] 官方的模型发布说明将此条目的日期标注为 2025 年 1 月 31 日
OpenAI 发布 o3-mini，一款面向编程、数学和科学的高性价比推理模型，提供三档可调推理强度。
0.66
软件2026/09/22@ScienceDesk
Codd Publishes A Relational Model of Data for Large Shared Data Banks
IBM Almaden Research Center, 650, Harry Road, IBM Almaden Research Center, San Jose, Santa Clara County, California, 95120, United StatesIBM
开始于 1970/06/30预估可信度 75%56.2年前The paper's own header gives the issue and no day: Communications of the ACM, "Volume 13 / Number 6 / June, 1970". A month with no day is estimated at that month's last day.^[1]
Ted Codd's paper in Communications of the ACM proposes storing data as relations queried by predicate logic, independent of how it is laid out on disk. It is the founding document of the relational database and of SQL.
0.65
AI2026/09/22@ScienceDesk
AlexNet Wins ImageNet and Starts the Deep Learning Boom
University of Toronto St. George, Chestnut Street, Yonge-Bay Corridor, University—Rosedale, Toronto, Golden Horseshoe, Ontario, M5G 1E2, CanadaUniversity of Toronto
开始于 2012/09/30已确认可信度 87%14年前Wikipedia's^[2][3] AlexNet article gives the submission date without hedging: "The three formed team SuperVision and submitted AlexNet in the ImageNet Large Scale Visual Recognition Challenge on September 30, 2012." The paper itself, "ImageNet Classification with Deep Convolutional Neural Networks", followed in Advances in Neural Information Processing Systems 25 later that year; the pin takes the dated competition result.
Alex Krizhevsky, Ilya Sutskever and Geoffrey Hinton submit a deep convolutional network to the ImageNet challenge as team SuperVision and win by a wide margin. The result turns GPU deep learning from a niche into an industry.
0.65
AI2026/09/19@TechDesk
GPT-5.1-Codex-Max 发布
OpenAI
开始于 2025/11/19 已确认可信度 90%308天前OpenAI^[1][2][3] 自家的模型发布说明将该条目的日期标注为 2025 年 11 月 19 日
OpenAI 发布 GPT-5.1-Codex-Max，这是一款前沿智能体编程模型，面向 Codex 中长时间运行的项目级任务。
0.65
AI2026/09/19@TechDesk
OpenAI 发布 o3 与 o4-mini
OpenAI
开始于 2025/04/16 已确认可信度 88%1.4年前OpenAI^[1] 自家的模型发布说明将该条目的日期标为 2025 年 4 月 16 日
OpenAI 发布其最强推理模型 o3，以及针对快速、低成本推理优化的小型模型 o4-mini。
0.64
AI2026/09/23@TechDesk
NVIDIA Vera Rubin NVL72 Debuts in MLPerf Inference v6.1
NVIDIA…NVIDIA Endeavor, 2788 San Tomas Expressway, Santa Clara, California 95051, United States
开始于 2026/09/16 15:00已确认可信度 86%7天前MLCommons^[4] published the MLPerf Inference v6.1 results on 16 September 2026 - its GlobeNewswire release is datelined 'SAN FRANCISCO, Sept. 16, 2026' and stamped '11:00 AM EDT' - and NVIDIA's^[1][2] post of the same day says the results were 'released today'.
In its first peer-reviewed benchmark, NVIDIA's Vera Rubin NVL72 rack posted up to 3.7x the throughput of GB300 NVL72 on Qwen3-VL and 2.5x on DeepSeek-R1, submitted in the preview category of MLCommons' MLPerf Inference v6.1 round.
0.64
AI2026/09/19@TechDesk
GPT-4o Update Brings More Natural Writing
OpenAI
开始于 2024/11/20 已确认可信度 95%1.8年前OpenAI's^[1] own model release notes date this entry November 20, 2024
OpenAI updates GPT-4o for all paid ChatGPT tiers with more natural, audience-aware writing and deeper work with uploaded files.
0.63
AI2026/09/19@TechDesk
GPT-5.3-Codex 发布
OpenAI
开始于 2026/02/05 已确认可信度 90%230天前OpenAI^[1][2][3] 官方的模型发布说明将该条目的日期标注为 2026 年 2 月 5 日
OpenAI 推出 GPT-5.3-Codex，这是其能力最强的智能体编程模型，也是首个融合 Codex 与 GPT-5 训练体系的模型。
0.63
AI2026/09/19@TechDesk
GPT-5-Codex Released
OpenAI
开始于 2025/09/15 已确认可信度 90%去年OpenAI's^[1][2][3] own model release notes date this entry Sep 15, 2025
OpenAI adds GPT-5-Codex, a GPT-5 variant optimized for agentic coding, across Codex.
0.63
AI2026/09/19@TechDesk
GPT-5.2 Thinking 恢复 Extended 思考级别
OpenAI
开始于 2026/02/04 已确认可信度 95%231天前OpenAI^[1] 官方的模型发布说明将该条目的日期标注为 2026 年 2 月 4 日
OpenAI 在 ChatGPT 中恢复了 GPT-5.2 Thinking 的 Extended 思考级别，撤销了 1 月误操作导致的下调。
0.63

Nature's^[1] own page for the article gives citation_online_date 2021/07/15 and shows "15 July 2021", and Wikipedia^[2][3] says the same: "On 15 July 2021, the AlphaFold 2 paper was published in Nature as an advance access publication alongside open source software and a searchable database of species proteomes." The print issue is dated August 2021.

开始: 2021年7月15日可信度 90%来自来源

DeepMind Publishes the AlphaFold 2 Paper

Nature publishes Highly accurate protein structure prediction with AlphaFold, alongside open-source code and a database of predicted proteomes. Protein structure prediction stops being a bottleneck for drug discovery.

"Highly accurate protein structure prediction with AlphaFold" was published in Nature on 15 July 2021, volume 596, pages 583-589, as an advance access publication^[1][2]
It came out alongside open source software and a searchable database of species proteomes^[2]
The lead author is John Jumper and the author list runs to more than thirty, most of them at DeepMind in London^[1]
The work built on AlphaFold 1, which placed first overall at the 13th Critical Assessment of Structure Prediction in December 2018^[2]
AlphaFold 3, co-developed with Isomorphic Labs, was announced on 8 May 2024^[2]

@ScienceDesk

参考资料 3可信度 87%

第一项始终是图钉的来源。总体可信度是各份资料对上方所用开始和结束时间支持程度的加权平均；资料每比最新的一份旧 180 天，权重减半。

[1]90%
nature.com/articles/s41586-021-03819-2
nature.com· 发布于 2026年9月22日· 开始 2021年7月15日 ✓· 占评分 33%
Nature's own page for the article gives citation_online_date 2021/07/15 and shows "15 July 2021", and Wikipedia^[2][3] says the same: "On 15 July 2021, the AlphaFold 2 paper was published in Nature as an advance access publication alongside open source software and a searchable database of species proteomes." The print issue is dated August 2021.
[2]90%
AlphaFold
en.wikipedia.org· 添加于 2026年9月22日· 占评分 33%
Dates the Nature^[1] paper and the open-sourcing, and gives the competition record behind it: AlphaFold 1 placed first at CASP13 in December 2018, "particularly successful at predicting the most accurate structures for targets rated as most difficult".
[3]80%
Protein structure prediction
en.wikipedia.org· 添加于 2026年9月22日· 占评分 33%
The problem the paper largely solved, for what was previously required - crystallography, cryo-EM and NMR - and why a computational answer changes the economics.^[1]

建议更正

有遗漏或错误吗？用你自己的话说明：能佐证此图钉的链接、不同的开始或结束日期及理由，或缺失、有误的信息。AI 会对照此图钉的来源进行核实，搜索更好的来源，并添加任何支持你说法的页面。图钉自身的来源仍然最重要。

串

回应这个图钉

1DeepMind Publishes the AlphaFold 2 Paper

AI2026/09/22@ScienceDesk
Google Publishes Attention Is All You Need
Googleplex, 1600, Amphitheatre Parkway, Mountain View, Santa Clara County, California, 94043, United StatesGoogle…
开始于 2017/06/12已确认可信度 88%9.3年前arXiv's^[1] own metadata for 1706.03762 gives citation_date 2017/06/12, and Wikipedia's^[2][3] article on the paper says the same: "On 2017-06-12, the original (100M-parameter) encoder-decoder transformer model was published in the 'Attention is all you need' paper."
Eight researchers at Google post a paper replacing recurrence and convolution with self-attention alone. The transformer architecture it introduces becomes the basis of nearly every large language model that follows.
0.67
AI2026/09/19@TechDesk
OpenAI Releases Open-Weight gpt-oss-120b and gpt-oss-20b
OpenAI
开始于 2025/08/05 已确认可信度 88%1.1年前OpenAI's^[1][2] own model release notes date this entry August 5, 2025
OpenAI releases two open-weight reasoning models, gpt-oss-120b and gpt-oss-20b, for teams that want to run and customize models themselves.
0.66
AI2026/09/19@TechDesk
OpenAI 发布 o3-mini
OpenAI
开始于 2025/01/31 已确认可信度 92%1.6年前OpenAI^[1] 官方的模型发布说明将此条目的日期标注为 2025 年 1 月 31 日
OpenAI 发布 o3-mini，一款面向编程、数学和科学的高性价比推理模型，提供三档可调推理强度。
0.66
软件2026/09/22@ScienceDesk
Codd Publishes A Relational Model of Data for Large Shared Data Banks
IBM Almaden Research Center, 650, Harry Road, IBM Almaden Research Center, San Jose, Santa Clara County, California, 95120, United StatesIBM
开始于 1970/06/30预估可信度 75%56.2年前The paper's own header gives the issue and no day: Communications of the ACM, "Volume 13 / Number 6 / June, 1970". A month with no day is estimated at that month's last day.^[1]
Ted Codd's paper in Communications of the ACM proposes storing data as relations queried by predicate logic, independent of how it is laid out on disk. It is the founding document of the relational database and of SQL.
0.65
AI2026/09/22@ScienceDesk
AlexNet Wins ImageNet and Starts the Deep Learning Boom
University of Toronto St. George, Chestnut Street, Yonge-Bay Corridor, University—Rosedale, Toronto, Golden Horseshoe, Ontario, M5G 1E2, CanadaUniversity of Toronto
开始于 2012/09/30已确认可信度 87%14年前Wikipedia's^[2][3] AlexNet article gives the submission date without hedging: "The three formed team SuperVision and submitted AlexNet in the ImageNet Large Scale Visual Recognition Challenge on September 30, 2012." The paper itself, "ImageNet Classification with Deep Convolutional Neural Networks", followed in Advances in Neural Information Processing Systems 25 later that year; the pin takes the dated competition result.
Alex Krizhevsky, Ilya Sutskever and Geoffrey Hinton submit a deep convolutional network to the ImageNet challenge as team SuperVision and win by a wide margin. The result turns GPU deep learning from a niche into an industry.
0.65
AI2026/09/19@TechDesk
GPT-5.1-Codex-Max 发布
OpenAI
开始于 2025/11/19 已确认可信度 90%308天前OpenAI^[1][2][3] 自家的模型发布说明将该条目的日期标注为 2025 年 11 月 19 日
OpenAI 发布 GPT-5.1-Codex-Max，这是一款前沿智能体编程模型，面向 Codex 中长时间运行的项目级任务。
0.65
AI2026/09/19@TechDesk
OpenAI 发布 o3 与 o4-mini
OpenAI
开始于 2025/04/16 已确认可信度 88%1.4年前OpenAI^[1] 自家的模型发布说明将该条目的日期标为 2025 年 4 月 16 日
OpenAI 发布其最强推理模型 o3，以及针对快速、低成本推理优化的小型模型 o4-mini。
0.64
AI2026/09/23@TechDesk
NVIDIA Vera Rubin NVL72 Debuts in MLPerf Inference v6.1
NVIDIA…NVIDIA Endeavor, 2788 San Tomas Expressway, Santa Clara, California 95051, United States
开始于 2026/09/16 15:00已确认可信度 86%7天前MLCommons^[4] published the MLPerf Inference v6.1 results on 16 September 2026 - its GlobeNewswire release is datelined 'SAN FRANCISCO, Sept. 16, 2026' and stamped '11:00 AM EDT' - and NVIDIA's^[1][2] post of the same day says the results were 'released today'.
In its first peer-reviewed benchmark, NVIDIA's Vera Rubin NVL72 rack posted up to 3.7x the throughput of GB300 NVL72 on Qwen3-VL and 2.5x on DeepSeek-R1, submitted in the preview category of MLCommons' MLPerf Inference v6.1 round.
0.64
AI2026/09/19@TechDesk
GPT-4o Update Brings More Natural Writing
OpenAI
开始于 2024/11/20 已确认可信度 95%1.8年前OpenAI's^[1] own model release notes date this entry November 20, 2024
OpenAI updates GPT-4o for all paid ChatGPT tiers with more natural, audience-aware writing and deeper work with uploaded files.
0.63
AI2026/09/19@TechDesk
GPT-5.3-Codex 发布
OpenAI
开始于 2026/02/05 已确认可信度 90%230天前OpenAI^[1][2][3] 官方的模型发布说明将该条目的日期标注为 2026 年 2 月 5 日
OpenAI 推出 GPT-5.3-Codex，这是其能力最强的智能体编程模型，也是首个融合 Codex 与 GPT-5 训练体系的模型。
0.63
AI2026/09/19@TechDesk
GPT-5-Codex Released
OpenAI
开始于 2025/09/15 已确认可信度 90%去年OpenAI's^[1][2][3] own model release notes date this entry Sep 15, 2025
OpenAI adds GPT-5-Codex, a GPT-5 variant optimized for agentic coding, across Codex.
0.63
AI2026/09/19@TechDesk
GPT-5.2 Thinking 恢复 Extended 思考级别
OpenAI
开始于 2026/02/04 已确认可信度 95%231天前OpenAI^[1] 官方的模型发布说明将该条目的日期标注为 2026 年 2 月 4 日
OpenAI 在 ChatGPT 中恢复了 GPT-5.2 Thinking 的 Extended 思考级别，撤销了 1 月误操作导致的下调。
0.63

相似图钉

参考资料 3可信度 87%

建议更正

评论

串

相似图钉