Akari Asai, Allen Institute for AI research scientist and incoming Carnegie Mellon assistant professor
Photo: Courtesy of Akari Asai via her official personal website; photographer not stated · Publisher-directed editorial display; source copyright retained

FigureAsia 35 Under 35 · Science

Akari Asai

Age 30 · Open retrieval and scientific language models · Japan / United States

First author of OpenScholar, an open retrieval-augmented system for evidence-grounded scientific synthesis.

Approximate age at the edition eligibility date
30
Field
Computer science
Country or region
Japan / United States
FigureAsia U35 Assessment
86.5 / 100

Career and documented record

Scientific question-answering is only useful when the literature behind an answer can be found and challenged. Akari Asai is first author of OpenScholar, published in Nature in 2026, an open retrieval-augmented system designed to search papers, synthesise evidence and expose its citations.

On ScholarQABench, a 2,967-question evaluation built with subject experts, the reported eight-billion-parameter system outperformed GPT-4o by 6.1 percentage points under the study's scoring. More important than that comparison is the architecture: retrieval, evidence selection and answer generation remain separable and inspectable.

OpenScholar cannot guarantee that every cited paper is correct or that synthesis is complete. It offers a stronger research contract: open models, explicit sources and a benchmark shaped by scientific users rather than generic web questions.

Why Akari Asai is on the list

Asai is selected for putting openness and evidence at the centre of scientific language modelling. OpenScholar answers a practical need while resisting the black-box form in which much AI assistance is sold.

The 2025–26 record

OpenScholar in Nature

First-authored an open retrieval-augmented system for scientific literature synthesis.

ScholarQABench

Evaluated on 2,967 expert-written scientific questions with citation-aware scoring.

Open research transition

Advanced the system at Ai2 while preparing an independent CMU faculty programme.

The work in its field

Scientific synthesis requires provenance as much as fluent text. Retrieval systems are strongest when users can see the documents, query the selection and distinguish evidence from generated connective prose.

Assessment breakdown

86.5out of 100

01

Substantive 2025–2026 contribution

16.7 / 20

First-authored an open retrieval-augmented system for scientific literature synthesis.

02

Verified scientific impact

12.5 / 15

Nature publication, open artefacts and a large expert-built benchmark give the work both visibility and practical research value.

03

Originality and distinction

8.8 / 10

The distinction lies in joining open retrieval, citation-grounded generation and expert scientific evaluation in one inspectable system.

04

Field influence

8.6 / 10

Researchers in open retrieval and scientific language models now have a stronger result to test, extend or challenge because of this contribution.

05

Individual agency

8.8 / 10

Asai is first author and a principal designer, with dataset, engineering and evaluation credit retained for the team.

06

Durability and trajectory

4.5 / 5

The contribution builds on an active line of work at Allen Institute for AI and Carnegie Mellon University, with further tests and applications still to come.

07

Asian significance and global relevance

4.5 / 5

Japanese computer scientist whose education and research career connect Japan with the United States.

08

Evidential validity and reproducibility

6.9 / 8

The benchmark is explicit and citation-aware; the profile does not treat model scores as factual infallibility.

09

Advance in scientific knowledge

6.2 / 7

The study offers evidence about how smaller open systems can compete in domain-specific scientific synthesis.

10

Translational or methodological utility

4.5 / 5

Researchers gain an open alternative for literature discovery and evidence-linked drafting.

11

Responsible research stewardship

4.5 / 5

Open models, visible citations and disclosed failure modes strengthen accountability in scientific assistance.

Evidence and attribution

Material claims on this page are supported by the edition’s evidence record. FigureAsia tests age, identity, role, result and individual attribution before publication. Public profiles present the reported record; supporting documentation is retained for accuracy review and corrections.

Achievement records
3
Assessment window
2025–26
Editorial status
Included in the 2026 FigureAsia 35 Under 35 edition

Rights and credit

The portrait is published under the rights basis recorded for this edition. Third-party ownership and reuse restrictions remain in force.

Publication status
Published under a documented rights basis
Credit
Courtesy of Akari Asai via her official personal website; photographer not stated
Licence
Publisher-directed editorial display; source copyright retained
Portrait source and credit