UT Arlington robotics researcher spanning construction robotics and embodied robot learning; early 2026 arXiv record
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
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—
This axis is not yet scored: the current snapshot has no verified publication-volume data for this researcher. It would measure total scholarly output — number of preprints and peer-reviewed papers, their venue types, and how the record grows over time.
Impact—
Not yet scored: citation counts and downstream influence are unavailable in the current snapshot. This axis would measure how often the work is cited and taken up by other studies, across papers and years.
Focus0/100
The focus score is 0 because the two cataloged 2026 preprints split the field distribution exactly evenly — 50% robotics (construction robotics and robot learning) and 50% robot learning (embodied AI / imitation learning). With only two papers, neither thread dominates yet; this reads as deliberate early breadth across an applications track (force-limited peg-in-hole insertion) and a learning track (vision-language-action policies) rather than established concentration.
Collaboration39/100
The collaboration score is 39, computed from 6 distinct coauthors across the two cataloged papers. Recurring collaborators include Lin He (listed twice in the co-author data), Yanyi Chen, Lingyao Li, Min Deng, and Hao Li. On both 2026 preprints Sun holds a middle author position and is not the corresponding author, indicating a contributing role within team-led projects rather than an independent, corresponding-author-led network.
Activity—
Not yet scored: the cataloged record spans only 2026, too short to characterize a publishing rhythm. This axis would measure output frequency, continuity, and gaps across multiple years of the record.
Reproducibility55/100
The reproducibility score is 55, a mean over the 2 papers with reproducibility assessments, both rated medium readiness. Evolve's verdicts credit honest controls and robustness diagnostics in the 0.8B geometry-conditioning study, while flagging thin hardware and gate validation in the SAGE insertion work, 3 evaluation seeds, and a weak 24.4% baseline — solid methodological transparency with clear room to strengthen the empirical support.
Research summary
Haofei Sun is a robotics researcher at The University of Texas at Arlington whose cataloged record splits evenly between two threads: construction robotics and embodied robot learning. Both entries are 2026 arXiv preprints with middle authorship and no corresponding role. The first, SAGE, couples LLM-based information extraction and retrieval-augmented generation over documented material strengths with a finite element response model to set force limits for robotic peg-in-hole insertion, cutting mean capacity error from 80.65% to 10.74%; Evolve notes hardware and gate validation remain thin. The second reports no reliable benefit from geometry conditioning in a 0.8B vision-language-action manipulation policy evaluated on LIBERO-Spatial and trained with gated delta networks and LoRA adaptation; Evolve credits honest controls and robustness diagnostics but flags 3 evaluation seeds and a weak 24.4% baseline. Bibliometrically, the record shows 6 distinct coauthors (collaboration score 39), including frequent collaborators Lin He, Yanyi Chen, Lingyao Li, Min Deng, and Hao Li. The field distribution is an exact 50/50 split, producing a focus score of 0, and reproducibility averages 55 across 2 assessed papers, both rated medium readiness. Output, impact, and activity remain unscored pending additional data.
Collaboration pattern
The collaboration score is 39, drawn from 6 distinct coauthors across the two cataloged papers. Recurring collaborators include Lin He (who appears twice in the co-author data), Yanyi Chen, Lingyao Li, Min Deng, and Hao Li. On both 2026 preprints Sun holds a middle author position and is not the corresponding author, indicating a contributing role in team-led projects rather than an independent, corresponding-author-led network.
Research keywords
Highlights
SAGE: from documented strengths to force limits for robotic insertion
A controls-first null result on geometry conditioning in a 0.8B embodied SLM
A two-track, diagnostics-heavy early record
Across both cataloged 2026 preprints, the work pairs an applications track (force-limited robotic insertion) with a learning-side diagnostics track (null-result controls in a small embodied policy), and both papers receive medium reproducibility ratings, averaging 55 over the 2 assessed papers.
Timeline
- 2026SAGE preprint posted on arXiv: material-informed robotic insertion for construction assembly, with middle authorship and no corresponding role.
- 2026Second preprint posted on arXiv: a geometry conditioning study in a 0.8B embodied SLM reporting no reliable benefit, again as a middle author.
- 2026Both 2026 preprints enter the Evolve catalog with medium reproducibility ratings (mean 55 over 2 assessed papers) and 6 distinct coauthors recorded.
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 signals come from a public OpenAlex-based snapshot; no ORCID is linked and 'Haofei Sun' is a common name, so identity is resolved from bibliographic data (affiliation and co-author overlap), not from verified biography.
- The cataloged record holds only two 2026 arXiv preprints with middle, non-corresponding authorship; peer-review status and final venues are unknown.
- Output, impact, and activity axes are unscored because no publication-count or citation data is available in this snapshot.
- The reproducibility mean of 55 is computed over only 2 assessed papers, so it is a small-sample figure that may change as more work is reviewed.