Embodied AI researcher at BAAI: dexterous VLA models, zero-shot 3D part segmentation, and motion reasoning
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
11
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: publication volume and authorship profile — total works, venue quality, first/corresponding-author share — cannot be computed because the current OpenAlex snapshot lacks sufficient validated output data for this axis.
Impact—
Not yet scored: citation impact — total citations, per-paper influence, and citation trajectory — is unavailable in the current snapshot, so no estimate is offered; the axis would measure how widely these works are cited and adopted.
Focus0/100
The focus score is 0 because the OpenAlex field distribution is perfectly even — one third each in robotics/robot learning, 3D perception, and motion reasoning — and five topics span from robot manipulation and learning (26) and reinforcement learning in robotics (23) to biomimetic flight (10). This indicates breadth across embodied AI rather than concentration in a single subfield.
Collaboration62/100
The collaboration score is 62, based on 22 distinct co-authors over the last three years and a stable recurring core — Yankai Fu, Ning Chen, Junkai Zhao and nine others appear repeatedly. All three catalogued 2026 papers list Yao as a middle author, suggesting contribution within larger team-led projects rather than solo or corresponding-author leadership.
Activity100/100
Activity scores 100: OpenAlex counts 33 works in the last three years, and counts-by-year shows 4 works in 2025 rising to 28 in 2026 — the strongest recent output level in the record. All three Evolve-reviewed preprints appear in 2026, consistent with this burst of productivity.
Reproducibility32/100
Reproducibility scores 32 (mean over 3 reviewed papers). LPA-CWM and DeCAL are rated low — no code stated, and DeCAL's real-robot gains come from 20-trial evaluations without significance tests; UniPart reaches medium with convincing ablations but mixed self-built benchmarks and single-run results. Stated code release would most directly raise this axis.
Research summary
Guaocai Yao is a researcher at the Beijing Academy of Artificial Intelligence working across robot learning and computer vision for embodied AI. OpenAlex bibliometrics attribute an h-index of 11 and an i10-index of 11, with 33 works in the last three years and 22 distinct co-authors; the annual record shows a sharp recent ramp — 4 works in 2025 and 28 in 2026 — against sparse earlier entries. Outputs split evenly across three areas: robotics (robot learning/embodied AI), 3D perception, and motion reasoning/point tracking, with topic concentrations in robot manipulation and learning (26) and reinforcement learning in robotics (23). Evolve's reviewed catalog includes three 2026 arXiv preprints where Yao is a middle author: DeCAL, a contact-aware dexterous vision-language-action model built on tactile gating and latent co-imagination; UniPart, a zero-shot language-grounded 3D part segmentation method with the LangPart-1M data recipe; and LPA-CWM, a learned adjudicator over counterfactual world model motion candidates (+60%/+29% DCAavg over uniform on DAVIS/Kinetics). Reproducibility is the current weak axis: a reviewed mean of 32 across 3 papers, with small-scale evaluations, no significance tests in one case, and code not stated for two of the three.
Collaboration pattern
Yao works within sizeable embodied-AI teams, appearing as a middle author on all three reviewed 2026 preprints. OpenAlex records 22 distinct co-authors over the last three years and a recurring core group — Yankai Fu, Ning Chen, Junkai Zhao, Heng Zhang, Pengwei Wang, Zhongyuan Wang, Shanghang Zhang, Xinqiang Yu, Zekun Qi, Jiawei He, Wenyao Zhang, and Xuchuan Chen — pointing to stable collaboration clusters around BAAI and partner labs rather than solo work.
Research keywords
Highlights
Contact-aware latent co-imagination for dexterous VLA models
Zero-shot language-grounded 3D part segmentation
Learned adjudicator reweights world-model motion candidates
Timeline
- 2026Released DeCAL, a contact-aware dexterous vision-language-action model combining tactile gating with latent co-imagination (arXiv preprint, middle author).
- 2026Contributed to UniPart on zero-shot language-grounded 3D part segmentation, introducing the LangPart-1M data recipe (arXiv preprint, middle author).
- 2026Co-authored LPA-CWM, a learned physical adjudicator for motion reasoning over counterfactual world model candidates (arXiv preprint, middle 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.
- Bibliometrics (h-index 11, i10-index 11, counts-by-year, topics, field distribution) come from an OpenAlex name-resolved profile built on public bibliographic data and may include same-name records.
- Output and impact dimensions are unscored: citation data in the current snapshot is insufficient, and no citation counts beyond the reported h/i10 indices are available.
- All three reviewed papers are 2026 arXiv preprints with middle-author positioning; peer-review status and final venues are unverified.
- The reproducibility axis reflects only the 3 papers reviewed by Evolve, not the full publication record.