Machine learning researcher at Nanyang Technological University focused on test-time adaptation and robustness
3
papers here
1
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
11
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) 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—
Total publication volume is not yet scored: the underlying comprehensive output count for this profile is insufficient to compute a reliable index. This axis would measure overall scholarly production — cumulative works, annual output, and venue spread across the career.
Impact—
Citation impact is not yet scored: aggregated citation counts and per-paper influence data are insufficient for a stable estimate. This axis would measure scholarly influence — citations, downstream uptake, and the visibility of this researcher's work across the field.
Focus100/100
The focus score of 100 reflects an unusually concentrated topic profile. Domain Adaptation and Few-Shot Learning (43 works) and Advanced Neural Network Applications (40) dominate the record, together outweighing the secondary clusters of Adversarial Robustness (17), Data Stream Mining (12), and Generative Adversarial Networks and Image Synthesis (9). The reviewed 2026 papers stay on-message: test-time adaptation, invariance learning under distribution shift, and vision-language spatial reasoning.
Collaboration64/100
A collaboration score of 64 sits in the middle range: the recent record lists 24 distinct coauthors, and 12 recur often, including Gang Dai, Gege Zhang, Lei Sun, Shuangping Huang, and Yaofo Chen. Across the three reviewed 2026 arXiv papers Niu appears as a middle author twice and first author once, suggesting steady team-based work rather than a solitary profile.
Activity100/100
The activity score of 100 marks an exceptionally high current tempo. Publication counts climb from 5 works in 2024 to 15 in 2025 and 21 in 2026, and the record logs 41 works over the most recent three-year window — a steep acceleration relative to the 2019-2024 stretch, when annual counts ranged from 1 to 7.
Reproducibility20/100
A reproducibility score of 20, averaged over one assessed paper, signals weak artifact availability. Evolve's verdicts point the same way: the geometry-aware test-time learning paper ships no code and relies on small test sets, while CATeye's core dataset remains private — so independent replication of this recent line of work is currently difficult.
Research summary
Shuaicheng Niu is a machine learning researcher at Nanyang Technological University whose OpenAlex-linked profile centers on test-time adaptation, domain generalization, and robust learning. The bibliometric record shows an accelerating output curve: from single-digit annual publication counts between 2019 and 2024 (1, 2, 5, 3, 7, and 5 works respectively) to 15 works in 2025 and 21 in 2026, with 41 works in the most recent three-year window. An h-index of 11 and i10-index of 12 indicate a citation footprint still consolidating relative to this output surge. Topic modeling of the record is dominated by Domain Adaptation and Few-Shot Learning (43 works) and Advanced Neural Network Applications (40), with secondary clusters in Adversarial Robustness (17), Data Stream Mining (12), and Generative Adversarial Networks and Image Synthesis (9). Among three 2026 arXiv papers reviewed by Evolve, Niu is first author of a survey unifying test-time adaptation, learning, and scaling under a feedback-driven test-time intelligence view, and a middle author on geometry-aware test-time LoRA adaptation for vision-language spatial reasoning and on CATeye, an invariance-learning method for zero-shot voucher abuse detection. Reproducibility signals are currently weak: the single assessed paper averages 20, reflecting missing code and private datasets.
Collaboration pattern
The recent three-year window lists 24 distinct coauthors, with a stable recurring core of 12 frequent collaborators including Gang Dai, Gege Zhang, Lei Sun, Shuangping Huang, Guohao Chen, and Yaofo Chen. Across the three Evolve-reviewed 2026 papers, Niu first-authors the test-time intelligence survey and contributes as a middle author on the VLM spatial reasoning and voucher abuse detection papers — a pattern of team-based collaboration with recurring partners rather than lone-author work.
Research keywords
Highlights
First-author survey unifying self-improving test-time intelligence
Geometry-aware test-time LoRA adaptation for VLM spatial reasoning
CATeye: invariance learning for zero-shot voucher abuse detection
Timeline
- 2019First publication year in the record, with 1 work.
- 2021Output rises to 5 works, a new high for the record to that point.
- 20237 works published — the strongest year before 2025.
- 2025Output accelerates to 15 works in a single year.
- 2026Breakout year with 21 recorded works, including three Evolve-reviewed arXiv papers: a first-authored test-time intelligence survey, geometry-aware test-time learning for VLM spatial reasoning, and CATeye for voucher abuse detection.
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 scores and counts are derived from OpenAlex-linked data; author identity is name-resolved from public bibliographic records.
- Output and impact dimensions are unscored because the underlying aggregate data is insufficient; h-index (11) and i10-index (12) are the only citation-side figures available.
- Reproducibility evidence covers only one assessed paper (mean score 20) and should not be read as a full career assessment.
- Qualitative highlights rest on the three 2026 arXiv papers reviewed by Evolve; other works appear only as bibliometric counts.