NUS researcher linking LLM agents for science and VLM-driven robot control across a broad multimodal AI agenda
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
29
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 yet scored: this snapshot lacks the raw output metrics — total publication volume, venue tiers and output trajectory — that this axis would measure. We make no productivity estimate; only the two reviewed 2026 preprints document verifiable output.
Impact—
Not yet scored: field-normalized citation impact data is insufficient in this snapshot. This axis would measure how widely work is adopted and cited relative to field baselines; basic indicators (h-index 29, i10-index 62) exist but no impact composite.
Focus0/100
Scores at the low end, reflecting broad topical coverage: OpenAlex topic weights span five clusters, led by Multimodal Machine Learning Applications (143) ahead of GANs and Image Synthesis (85), Advanced Neural Network Applications (69), Domain Adaptation and Few-Shot Learning (50) and Human Pose and Action Recognition (38); the two reviewed 2026 papers split evenly between NLP for science and robotics (0.5/0.5 field distribution).
Collaboration64/100
Moderate-to-strong collaborative footprint: 24 distinct co-authors appear in the recent three-year window alongside 167 works, with 12 recurring collaborators — including Kevin Qinghong Lin, Pan Lu and Junchi Yu — returning across projects. On the two reviewed 2026 papers he sits in middle and last author positions without corresponding status, consistent with a co-lead or senior-collaborator role inside multi-person teams rather than solo output.
Activity100/100
Top-of-scale activity: 167 works in the recent three-year window, with OpenAlex yearly counts climbing from 12 (2021) and 19 (2022) to 56 (2023) and a peak of 62 (2024), followed by 59 (2025) and 46 already logged in 2026 — sustained high output across the last four years, including both reviewed 2026 preprints.
Reproducibility55/100
Middle-of-scale reproducibility (mean 55 across the 2 papers Evolve has assessed). Both 2026 preprints carry medium repro-readiness: PaperDoctor reports 70.6% author acceptance of its feedback but no baseline or independent correctness validation, while Show-Harness's 89% vs 57% zero-shot comparison lacks significance tests and latency reporting — evidence-backed claims, but validation gaps remain.
Research summary
Mike Zheng Shou is a researcher at the National University of Singapore whose recent work spans two equally weighted directions — natural language processing for science and robotics/embodied AI (0.5/0.5 in the OpenAlex field distribution). The bibliometric record shows an h-index of 29, an i10-index of 62, and 167 works in the recent three-year window, with yearly counts rising from 12 (2021) and 19 (2022) to 56 (2023), 62 (2024), 59 (2025) and 46 logged in 2026. Topic mass concentrates in Multimodal Machine Learning Applications (143), followed by GANs and Image Synthesis (85), Advanced Neural Network Applications (69), Domain Adaptation and Few-Shot Learning (50), and Human Pose and Action Recognition (38) — an unusually broad spread that explains the low focus score. The two Evolve-reviewed 2026 preprints illustrate the current agenda: PaperDoctor, a three-level LLM agent for evidence-grounded pre-submission feedback (70.6% author acceptance, no baseline validation), and Show-Harness, a VLM agent for zero-shot robot control (89% vs 57% best baseline, small-N). Both papers place him in middle and last author positions without corresponding status, amid 24 distinct co-authors and 12 recurring collaborators. Reproducibility scores mid-range (55) across the two assessed papers.
Collaboration pattern
The 24 distinct co-authors in the recent three-year window and 12 recurring collaborators — including Kevin Qinghong Lin, Pan Lu, Junchi Yu, Yupeng Chen and Sheng Liu — point to steady work inside a stable multi-person group. On the two reviewed 2026 papers he holds middle and last author positions without corresponding status, suggesting he contributes as co-lead or senior collaborator on team projects rather than as the sole corresponding driver.
Research keywords
Highlights
Three-level LLM agent gives evidence-grounded feedback on papers in progress
Semantic action units let a VLM agent drive robots zero-shot
Timeline
- 202112 OpenAlex-indexed works — the earliest year in the snapshot's counts.
- 202219 works; output still modest before the 2023 step change.
- 202356 works — output nearly triples year on year, marking a clear step change.
- 202462 works, the peak year in the snapshot.
- 202559 works — productivity held near peak.
- 202646 works logged in the snapshot; both reviewed preprints (PaperDoctor, Show-Harness) appear on arXiv.
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 indicators (h-index, yearly counts, topics, co-author counts) come from the OpenAlex identity record; no Scopus or Web of Science cross-check was performed.
- Output and impact dimensions are unscored because the underlying raw metrics are absent from this snapshot.
- The 2026 count (46 works) covers a partial calendar year in the snapshot.
- Reviewed evidence covers only two 2026 arXiv preprints; Evolve verdicts are independent assessments, not peer review.