Zhejiang University researcher in data-efficient language modeling and evaluation of autonomous research agents
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
37
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—
Not scored yet: this snapshot provides no verified per-year publication list for this identity, so output volume and authorship-role distribution cannot be measured; the axis would count works per year and the share of first or corresponding-author positions.
Impact—
Not scored yet: no citation totals, citing works or field-weighted citation measures are available in this snapshot, so scholarly influence cannot be assessed; the axis would track cumulative citations and citation impact relative to field norms.
Focus0/100
The focus score is 0, signaling a strongly dispersed topic profile. OpenAlex clusters 58 works in metamaterials and metasurfaces, 53 in topological materials, 37 in non-Hermitian quantum physics, 18 in acoustics and 13 in sarcoma, while the reviewed portfolio splits evenly (0.5/0.5) between data-efficient language modeling and agent evaluation.
Collaboration63/100
Collaboration scores 63. The recent three-year window contains 44 works with 23 distinct coauthors, and twelve recurring collaborators appear in the data, led by Kaihao Zhu, Taowen Deng, Shenzhan Hong and Zichen Li. Both reviewed 2026 papers list the researcher as last (non-corresponding) author, consistent with a supervisory role in team-based AI benchmarking work.
Activity100/100
Activity scores 100, the maximum. Yearly counts rise from 17 works in 2020 to 23 in 2021, 29 in 2023 and 30 in 2024, the highest annual total, with 44 works in the recent three-year window and 2025-2026 counts (10 and 4) still accruing. Output pace has intensified steadily since 2020.
Reproducibility78/100
Reproducibility scores 78, computed from the two Evolve-reviewed papers (reproMean 78). The BabyLM submission carries high reproducibility readiness thanks to careful controls, while the Qiushi Engine report is rated medium because attribution and judge validity rest on self-reported evidence; the signal is solid but drawn from a small sample.
Research summary
Yihao Yang (Zhejiang University, China; ORCID 0000-0002-9229-0054) is profiled here from name-resolved public bibliographic data plus two Evolve-reviewed papers. The reviewed work targets data-efficient language modeling and the evaluation of autonomous scientific research systems, with the researcher listed as last author on both 2026 arXiv papers. The first, a BabyLM 2026 submission built on restatements and data organization, tops the 8 September 2026 Strict-Small snapshot at 42.25 with careful controls, though Evolve notes the gains hinge on a 203-item near-chance metric and two seeds. The second reports the Qiushi Engine at 0.816 on AstaBench E2E-Bench-Hard, a 3.3x best perfect-task rate, while attributing gains through disciplined but self-reported rubric-based evaluation. Underlying bibliometrics under the OpenAlex identity show an h-index of 37, an i10-index of 68, and yearly counts climbing from 17 works in 2020 to 30 in 2024, with 44 works in the recent three-year window and 23 distinct coauthors. A focus score of 0 reflects topics dispersed across metamaterials, topological physics, acoustics and other clusters that diverge from the reviewed AI portfolio, so partial identity conflation cannot be excluded. Reproducibility averages 78 across the two reviewed papers.
Collaboration pattern
Collaboration scores 63. The recent three-year window spans 44 works with 23 distinct coauthors, and twelve recurring collaborators appear in the data: Kaihao Zhu, Taowen Deng, Shenzhan Hong, Zichen Li, Jincheng Mi, Yuang Pan, Hongsheng Chen, Shuxing Yang, Junjie Yang, Rui Zhao, Junyao Wu and Yize Wang. Both reviewed 2026 papers list the researcher as last (non-corresponding) author, consistent with a senior supervisory role in team-based AI benchmarking work.
Research keywords
Highlights
Tops BabyLM 2026 Strict-Small at 42.25 with data-centric pretraining
Qiushi Engine reports 0.816 on AstaBench E2E-Bench-Hard
Reproducibility discipline across the reviewed portfolio
Timeline
- 2010Earliest signal in the OpenAlex yearly counts: 1 recorded work, the starting point of this name-resolved snapshot.
- 2020Annual output climbed to 17 works, beginning a sustained acceleration from single-digit yearly levels (1-8 works, 2010-2019).
- 2024Highest annual count in the snapshot: 30 works, following 29 in 2023 — two consecutive peak-output years.
- 2026Two papers enter the Evolve catalog: the BabyLM 2026 Strict-Small program and the Qiushi Engine AstaBench report, both listing the researcher as last author.
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.
- Bibliometric metrics (h-index 37, i10-index 68, yearly counts, topics) come from an OpenAlex name-resolved identity; 'Yihao Yang' is a common name and resolution confidence is limited.
- The OpenAlex topic mix (metamaterials, topological and non-Hermitian physics, acoustics, sarcoma) diverges sharply from the reviewed NLP/AI-agent papers, so partial profile conflation cannot be ruled out.
- Output and impact dimension scores are missing because no per-paper citation data was provided; these axes are intentionally not estimated.
- 2025-2026 annual counts (10 and 4) likely reflect an incomplete indexing window rather than a decline in output.