Google researcher uniting mathematical programming, recommender retrieval, and kernel-based learning theory
2
papers here
2
reviewed by Evolve
1
first-author
1
corresponding
—
works · OpenAlex
6
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
Unscored: the output axis requires a complete, validated publication list, and the OpenAlex record on file is partial. Counts by year exist, but Evolve has not yet assembled the full corpus this axis measures, so no score is given.
Impact—
Unscored: impact would measure citations, uptake, and influence of the work. The only signals on file are thin OpenAlex counts (h-index 6, i10-index 4), too limited to support a defensible impact score.
Focus0/100
The focus score is 0, the lowest tier, because this record is broad rather than concentrated. The OpenAlex field distribution splits exactly 50/50 between machine learning theory and operations research / optimization in machine learning, and five topic clusters coexist: recommender systems (18 works), information retrieval and search behavior (14), multi-objective optimization (12), domain adaptation and few-shot learning (6), and risk and portfolio optimization (6).
Collaboration28/100
The collaboration score is 28, reflecting a deliberately small network: only 3 distinct co-authors appear in the last three years, with Zhe Li, Bicheng Ying, and Haibo Yang as the recurring partners. The two 2026 catalog papers fit the pattern — a first-authored, corresponding-author taxonomy paper, and a middle-author position on the kernel-based subliminal learning paper.
Activity84/100
The activity score is 84, the strongest axis in this profile. The record shows 17 works in the last three years, and 2026 already has 10 works — the highest annual figure in the counts-by-year series (2018: 2; 2020: 6; 2021: 4; 2022: 3; 2023: 9; 2024: 3; 2025: 4), a sharp rebound after the 2024-2025 dip.
Reproducibility38/100
The reproducibility score is 38, the mean over the 2 papers Evolve has reviewed, and the two diverge: the kernel-based subliminal learning paper carries medium readiness, with its theory validated on MLPs and Qwen2.5-7B, while the mathematical programming taxonomy is rated low because it is wholly non-empirical, offering mapping and attribution rather than runnable experiments.
Research summary
Chaosheng Dong is a researcher at Google (United States) whose public bibliometric record (OpenAlex, ORCID 0000-0003-4491-0594) spans two fields almost evenly: machine learning theory and operations research / optimization in machine learning (0.5 each). The topic profile is led by recommender systems and techniques (18 works), information retrieval and search behavior (14), and multi-objective optimization (12), with smaller clusters in domain adaptation and few-shot learning (6) and risk and portfolio optimization (6). Output has accelerated sharply: 17 works in the last three years, including 10 works already recorded for 2026 — the highest annual count in the record — after dips to 3 (2024) and 4 (2025). Evolve has reviewed two 2026 arXiv papers: a first-authored, corresponding-author taxonomy mapping 30 ML/AI applications onto ten optimization paradigms, judged conceptually solid but wholly non-empirical; and a middle-author kernel-based theory of subliminal learning, validated on MLPs and Qwen2.5-7B but idealized. Citation signals are thin (h-index 6, i10-index 4), so the output and impact axes remain unscored. Collaboration is narrow — only 3 distinct recent co-authors, with Zhe Li, Bicheng Ying, and Haibo Yang recurring — and reproducibility averages 38 across the two reviewed papers.
Collaboration pattern
The OpenAlex record lists only 3 distinct co-authors over the last three years, yielding a low collaboration score of 28. Zhe Li, Bicheng Ying, and Haibo Yang recur as the principal partners. Recent authorship roles alternate between leadership — first and corresponding author on the 2026 mathematical programming taxonomy — and a supporting middle position on the kernel-based subliminal learning paper, suggesting a compact, stable working circle rather than broad external networking.
Research keywords
Highlights
A kernel-level theory of subliminal learning
First-authored taxonomy of mathematical programming in ML and AI
Record-setting 2026 output surge
The counts-by-year record shows 10 works already logged for 2026 — the highest annual figure on file, ahead of the 2023 peak of 9 — within 17 works over the last three years; two of the year's papers entered the Evolve catalog.
Timeline
- 2018Earliest annual output in the OpenAlex counts-by-year record: 2 works logged.
- 2020Output rose to 6 works, an early uptick in the counts-by-year series.
- 2023Output reached 9 works, then the highest annual count on record.
- 2024Output dipped to 3 works in 2024 and recovered only partially to 4 in 2025.
- 2026Record-high 10 works logged so far; two arXiv papers entered the Evolve catalog — a first-authored, corresponding-author mathematical programming taxonomy, and a middle-author kernel-based theory of subliminal learning.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- All bibliometrics come from OpenAlex only; no citation data from Google Scholar, Scopus, or Web of Science was used.
- The output and impact axes are unscored because the citation record is thin (h-index 6, i10-index 4) and may undercount if the OpenAlex profile is incomplete.
- This profile is name-resolved from public bibliographic data (OpenAlex identity plus ORCID 0000-0003-4491-0594); 'Chaosheng Dong' is a fairly common name, and records were matched to the Google (US) affiliation.
- Reproducibility scoring covers only the 2 papers Evolve has reviewed, and 2019 is absent from the counts-by-year record, so coverage may be incomplete.