AI ChronicleAI 纪年 Full timeline →

Research breakthrough · curated entry

BERT refreshes NLP benchmarks with bidirectional pretraining

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

Date
October 11, 2018
Category
Research breakthrough
Organization / Model
Google AI Language
Evidence
1 source · 1 evidence family

Google released BERT, using a bidirectional Transformer and masked language modelling to reset multiple NLP benchmarks. Together with GPT it established the pretrain-then-fine-tune paradigm.

Curator's note: Cross-checked against the arXiv paper.

Curated sources · 1

  1. Paper preprints · arXiv2018-10-11

    BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding

    arxiv.org

Cite this page

This page is a curated static snapshot and can be cited directly. Preferred format GB/T 7714—2025 (electronic resource [EB/OL]); the other 5 styles open below, together with the machine-readable files.

GB/T 7714—2025 (electronic resource [EB/OL])

AI 纪年. BERT 发布，双向预训练刷新 NLP 基准[EB/OL]. (2018-10-11)[2026-11-03]. https://aihist.net/e/bert-2018.html.

APA 7 · American Psychological Association

AI 纪年. (2018, October 11). BERT 发布，双向预训练刷新 NLP 基准 [Web page]. https://aihist.net/e/bert-2018.html

MLA 9 · Modern Language Association

"BERT 发布，双向预训练刷新 NLP 基准." AI 纪年, 11 Oct. 2018, aihist.net/e/bert-2018.html. Accessed 3 Nov. 2026.

Chicago 17 · notes-bibliography

AI 纪年. "BERT 发布，双向预训练刷新 NLP 基准." Published Oct. 11, 2018. Accessed Nov. 3, 2026. https://aihist.net/e/bert-2018.html.

IEEE · citation-reference list

"BERT 发布，双向预训练刷新 NLP 基准," AI 纪年, 11 Oct. 2018. [Online]. Available: https://aihist.net/e/bert-2018.html. [Accessed: 3 Nov. 2026].

ACM · author-date variant

AI 纪年. October 11, 2018. BERT 发布，双向预训练刷新 NLP 基准. Retrieved November 3, 2026 from https://aihist.net/e/bert-2018.html.

All six styles come from one template and can be compared word for word. [引用日期], Accessed, urldate and Y2 all carry this page's build date, 2026-11-03 — a static page is an immutable snapshot, so substitute the date you actually consulted it. The address in every style is the archive's canonical page: this English page translates the same snapshot, and the work keeps its name, so one record cites both. For one-click copying use the same entry in the timeline, which takes the day you opened it. Expand the other styles before printing.