Institutional dossier
The documented history of Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院: chronology, people, research, capital and compute, and governance. Every statement opens its source and its limits.
Records dated 2018-11-14 to 2024-11-05 · 1 undated
40 claims · 17 sources · 40 not editorially accepted. The counts describe this selection. They do not measure completeness or independent confirmation.
Scope, evidence and limits
Description: Beijing AI research institute that describes itself as a non-profit research institution (非营利性科研机构) established in November 2018; developer of the Wu Dao (悟道), Aquila, Emu and BGE models. Its registered legal form was not retrieved.
The records are grouped automatically. No human inspection of the selected claims, source paths or framing is recorded for this dossier. Each source passage shows how it was checked.
The dossier is computed from existing records. It is not a continuous history of the institution, and a missing record does not mean that an event or relationship did not happen. The chronology uses the dates of events and reports, not the date the public could first have known of them. Later sources can describe earlier events.
Select any statement to see its exact claim, qualification, date note, evidence passage, location in the source, source version and review status. The description above is Selekoa’s stored description of the institution, based on these sources:
Chronology
Records about the institution itself and documented changes to its frameworks, by year. Expand a year to read the claims. People’s other institutional records and related documents have their own sections. Where a date gives only the year or month, the order within it is unknown.
2018Record year · 1 claims
- BAAI states that it was established on 14 November 2018; its second-anniversary post says it was set up under the guidance and support of the Ministry of Science and Technology and the Beijing municipal Party committee and government, promoted by the Beijing Municipal Science and Technology Commission and the Haidian District government.Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院 → is stated to have been founded in → Established on 14 November 2018 (as BAAI states it) Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
The sponsoring bodies are recorded in BAAI's own words (指导和支持, 推动成立: 'guidance and support', 'promoted the establishment'); no direction, control or ownership is inferred from them. The website describes BAAI as a non-profit research institution (非营利性科研机构); its registered legal form and registration record were not retrieved, so no has_legal_form record is proposed.
Day as given in BAAI's second-anniversary post of November 2020; BAAI's current website gives the month only (2018年11月).
2019Record year · 1 claims
- BAAI led the release of the Beijing AI Principles (人工智能北京共识) on 25 May 2019, together with universities, research institutes and an industry alliance.Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院 → published → Beijing AI Principles (人工智能北京共识)Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
The English text is hosted by the International Research Center for AI Ethics and Governance, not by BAAI. BAAI's own former page for the principles now returns its single-page-application shell; the cross-family check (2026-10-31) located the Chinese original (人工智能北京共识) on BAAI's site in an Internet Archive capture of 11 August 2020, which dates the page 25 May 2019, and it is recorded as a second expression. That the release was joint with other institutions rests on the English note and on China Daily's report. 发布 is rendered 'released'.
Release date as given in the 'Release and Endorsement' note of the English text and in China Daily's report of 26 May 2019.
2020Record year · 2 claims
- BAAI, with Peking University's and Tsinghua University's AI institutes, three Chinese Academy of Sciences institutes, several companies and an industry alliance, released the Beijing AI Principles for Children (面向儿童的人工智能北京共识) on 14 September 2020, according to BAAI's community post.Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院 → published → Beijing AI Principles for Children (面向儿童的人工智能北京共识)Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
The co-publishers are those the post names (including Xiaomi, Megvii and Qihoo 360); the post's description of the text as the country's earliest such principles (我国首个) is BAAI's wording and is not adopted. 共同发布 is rendered 'jointly released'.
Release date as stated in the post of 16 September 2020.
- BAAI's second-anniversary post names Huang Tiejun as president (院长) of BAAI at the 2020 BAAI Forum on 14 November 2020.Huang Tiejun → held role at → Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
院长 is rendered 'president' (also translated 'dean'). The post does not state when the role began or ended; it is not extended beyond the snapshot.
Role as shown on the date of the forum reported in the post; start and end dates are not stated.
2021Record year · 3 claims
- BAAI's account of the 2021 BAAI Conference names Tang Jie as BAAI's vice president for academic affairs (学术副院长), who presented Wu Dao 2.0 at the opening on 1 June 2021, and says he led the Wu Dao team.Tang Jie / 唐杰 → held role at → Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
学术副院长 is rendered 'vice president for academic affairs'. The role at BAAI is separate from Tang Jie's Tsinghua and Zhipu roles recorded elsewhere; no relation between them is inferred.
Role as shown at the conference opening of 1 June 2021; start and end dates are not stated.
- BAAI announced the Wu Dao 2.0 (悟道2.0) model at the opening of the 2021 BAAI Conference on 1 June 2021, according to BAAI's community post of 7 June 2021.Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院 → published → Wu Dao 2.0 release announcement (BAAI Conference, 1 June 2021)Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
The post's descriptions of Wu Dao 2.0 as the world's largest model and of Wu Dao 1.0 as China's earliest such model are BAAI's own wording and are not adopted; parameter counts and performance are out of scope. No weights or model card for Wu Dao 2.0 were inspected.
Date of the conference opening as stated in the post; the post itself is dated 2021-06-07. The same post says Wu Dao 1.0 was released on 20 March (2021).
- BAAI's account of the 2021 BAAI Conference names Zhang Hongjiang as chairman of BAAI's council (理事长) at the opening ceremony on 1 June 2021.Zhang Hongjiang → held role at → Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
理事长 is rendered 'chairman of the council'. The post does not state when the role began or ended.
Role as shown at the conference opening of 1 June 2021; start and end dates are not stated.
2023Record year · 5 claims
- BAAI published the Aquila-7B (悟道·天鹰) model card in its Hugging Face repository by 8 June 2023, describing a Chinese-English language model trained from scratch.Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院 → published → Aquila-7B model cardInspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
The card's claims of efficiency, performance and of being the earliest model with these features are the developer's own and are out of scope. The card at this commit still carries placeholder table values (0.xxx), indicating it was a working version.
Date of the cited README commit; the repository's commit list begins on 8 June 2023. BAAI's own website places Wu Dao 3.0, including Aquila, in 2023 without a day; the release event date was not read from a BAAI page.
- The Aquila-7B model card of 8 June 2023 states that the model is licensed under the BAAI Aquila Model Licence Agreement (智源Aquila系列模型许可协议) and the source code under Apache Licence 2.0.Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院 → published → Aquila-7B licence statement (BAAI Aquila Model Licence Agreement), commit c0687c52Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
The card's model-details table at the same commit lists 'Apache 2.0' in its License column for Aquila-7B, which differs from the licence section; the record follows the licence section and notes the discrepancy. The agreement's own text was not inspected, so its terms are not described. 开源 is rendered 'open-source' as the card's own label.
Date of the cited README commit; the agreement's PDF was uploaded to the repository the same day.
- Researchers affiliated with BAAI, Tsinghua University and Peking University posted 'Generative Pretraining in Multimodality', introducing the Emu multimodal foundation model, to arXiv on 11 July 2023 (version 1), with code under the baaivision GitHub organisation.Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院 → published → Generative Pretraining in Multimodality (Emu)Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
BAAI is affiliation number 1 and most authors are marked with it; attribution of the publication to BAAI rests on that and on the baaivision code link, not on a BAAI announcement. Performance claims are out of scope.
Submission date of the cited arXiv version; not the model's release date and not the date of later versions.
- The BAAI/bge-large-zh-v1.5 model card, at its commit of 12 September 2023, declares the MIT licence and states that the released models can be used commercially free of charge.Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院 → published → bge-large-zh-v1.5 licence statement (MIT License), commit bc5e88a5Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
The licence section names the MIT License for FlagEmbedding and adds the commercial-use sentence; the LICENSE file it links on GitHub was not inspected. Contact e-mail addresses in the card are not reproduced.
Date of the cited README commit, the day the repository was initialised; the card's own update list dates the release of the bge-*-v1.5 models to 09/12/2023.
- Researchers at BAAI, with a co-author at Hugging Face, posted 'C-Pack: Packaged Resources To Advance General Chinese Embedding' to arXiv on 14 September 2023 (version 1), describing the C-MTEB benchmark, the C-MTP dataset and the C-TEM family of embedding models.Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院 → published → C-Pack: Packaged Resources To Advance General Chinese EmbeddingInspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
Affiliations as printed on the title page; the C-TEM models are published on Hugging Face as BGE (recorded separately). Performance claims are out of scope.
Submission date of the cited arXiv version; not the model's release date and not the date of later versions.
2024Record year · 6 claims
- BAAI's Language Foundation Model & Software Team posted the Aquila2 Technical Report to arXiv on 14 August 2024 (version 1), describing bilingual models of 7, 34 and 70 billion parameters.Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院 → published → Aquila2 Technical ReportInspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
Affiliation as printed on the title page. Performance claims are out of scope.
Submission date of the cited arXiv version; not the model's release date and not the date of later versions.
- The BAAI/Emu3-Gen model card, at the commit of 26 September 2024, declares the licence apache-2.0 in its metadata.Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院 → published → Emu3-Gen licence statement (apache-2.0), commit 6f85fd46Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
The statement is the card's metadata field only; no LICENSE file in the repository was inspected, and the licences of the other Emu3 repositories (for example Emu3-Chat and the vision tokenizer) were not checked.
Date of the cited README commit; the repository was initialised on 25 September 2024.
- The Emu3 Team of BAAI posted 'Emu3: Next-Token Prediction is All You Need' to arXiv on 27 September 2024 (version 1).Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院 → published → Emu3: Next-Token Prediction is All You NeedInspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
Credit as printed on the title page. The timeline entry's description of Emu3 as the world's earliest native multimodal world model (全球首个) is BAAI's wording and is not adopted. The report's performance claims are out of scope.
Submission date of the cited arXiv version; not the model's release date and not the date of later versions. BAAI's website timeline places the Emu3 release in October 2024 (2024年10月).
- A BAAI news item dated 8 November 2024 names Huang Tiejun as chairman of BAAI's council (理事长) at a BAAI summit held on 5 November 2024.Huang Tiejun → held role at → Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
理事长 is rendered 'chairman of the council'. The change from the 2020 role (院长) to this one is shown only by two dated snapshots; no transition date is recorded.
Role as shown at the event of 5 November 2024, in a news item dated 2024.11.08 on BAAI's website; start and end dates are not stated.
- A BAAI news item dated 8 November 2024 names Lin Yonghua as a vice president (副院长) of BAAI at a BAAI summit held on 5 November 2024.Lin Yonghua → held role at → Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
副院长 is rendered 'vice president'. The same passage supports another record of the same event.
Role as shown at the event of 5 November 2024, in a news item dated 2024.11.08 on BAAI's website; start and end dates are not stated.
- A BAAI news item dated 8 November 2024 names Wang Zhongyuan as president (院长) of BAAI at a BAAI summit held on 5 November 2024.Wang Zhongyuan → held role at → Beijing Academy of Artificial Intelligence / 北京智源人工智能研究院Inspect claim & evidence ↗
Record about this institution · Stated in source · Supported
Qualification and date notes
院长 is rendered 'president'. The item does not state when the role began.
Role as shown at the event of 5 November 2024, in a news item dated 2024.11.08 on BAAI's website; start and end dates are not stated.