Robot learning researcher working on latent world models and robust calibration for vision-language-action policies.
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-13) 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—
No output score is available: the snapshot contains no publication counts, venue list, or preprint totals beyond the two reviewed 2026 arXiv papers. This axis would measure total and annual research output across venues and preprint servers.
Impact—
No impact score is available: the data provides no citation counts, h-index, or downstream adoption signals. This axis would measure how often the work is cited and built upon by other researchers and systems.
Focus0/100
Evolve's focus score is 0, computed from the field distribution across cataloged papers: the two reviewed 2026 preprints split evenly (0.5/0.5) between robotics/embodied machine learning and robot learning/world models (machine learning for robotics). With only two papers on file, this even split likely reflects coverage limits as much as genuine breadth, since the two areas are methodologically adjacent — both concern learned policies and world models for manipulation.
Collaboration55/100
The collaboration score is 55, based on 15 distinct coauthors across the two cataloged 2026 papers. The frequency list shows a recurring core — Shuaijun Liu, Chengyu Wu, and Feiyang You each appear twice — alongside Chenglong Zhang, Ningxin Su, Jingyao Cai, Xingwei Chen, Qifu Wen, and Xi Lin. On both papers Hao sits in a middle authorship position without corresponding status, consistent with specialist contributions inside moderately sized teams rather than solo or senior-led work.
Activity—
No activity score is available: with only two papers, both dated 2026, there is no multi-year publication cadence to compute a recent-output trend. This axis would measure publication frequency and research pace over recent years.
Reproducibility55/100
Reproducibility scores 55, averaged over two scored papers (reproMean 55, reproPapers 2), both rated medium readiness. For D-JEPA, Evolve notes consistent simulation and real-robot gains but flags cohort selection, a tiny driving set, and no released code. BAS-VLA shows solid simulation evidence, yet its robustness claim rests on narrow preserving evidence, including a 0% stale-success rate under semantic change. Code release and broader benchmarks would lift this axis.
Research summary
Shuyang Hao is an early-stage researcher in robot learning whose reviewed work centers on making learned robot policies more reliable at decision time. The Evolve catalog holds two 2026 arXiv preprints, both with Hao in a middle authorship position and neither corresponding. D-JEPA introduces a decision-aligned latent world model that re-ranks candidate latent futures with a bounded set transformer to close a decision-local gap; Evolve's review credits consistent simulation and real-robot gains while flagging cohort selection, a tiny driving set, and no released code. BAS-VLA adds a task-semantic action-calibration layer on frozen vision-language-action models: it remains stable under appearance shifts but collapses to 0% stale success under semantic change, with solid simulations but narrow preserving evidence. The field distribution splits evenly at 0.5/0.5 between robotics/embodied machine learning and robot learning/world models, which explains the focus score of 0 given only two cataloged papers. Collaboration scores 55, built on 15 distinct coauthors with recurring partners including Shuaijun Liu, Chengyu Wu, and Feiyang You; reproducibility also scores 55 across two medium-readiness papers. With output, impact, and activity metrics unavailable, this profile should be read as an early-stage picture drawn from a small curated sample rather than a complete publication record.
Collaboration pattern
Across the two cataloged 2026 papers, Hao works with 15 distinct coauthors (collaboration score 55). The frequency list highlights a stable core — Shuaijun Liu, Chengyu Wu, and Feiyang You each appear twice — plus Chenglong Zhang, Ningxin Su, Jingyao Cai, Xingwei Chen, Qifu Wen, and Xi Lin. Holding middle, non-corresponding positions on both preprints suggests Hao contributes as a technical specialist within established teams rather than leading them, at least in the reviewed sample.
Research keywords
Highlights
D-JEPA re-ranks latent-world-model futures for decision alignment
BAS-VLA separates appearance robustness from semantic fragility
An evaluation-first working style with medium reproducibility readiness
Both cataloged 2026 preprints are rated medium (reproMean 55, reproPapers 2) and share a controlled-intervention mindset — counterfactual evaluation in BAS-VLA, bounded candidate re-ranking in D-JEPA — but neither released code, the main lever for improving this axis.
Timeline
- 2026First cataloged preprint: D-JEPA, a decision-aligned latent world model re-ranking candidate futures with a bounded set transformer, appears on arXiv with Hao as a middle author.
- 2026Second cataloged preprint: BAS-VLA, a task-semantic action-calibration layer for frozen vision-language-action models, also appears on arXiv, again in a middle, non-corresponding position.
- 2026Within the same year, the two reviewed papers connect Hao to 15 distinct coauthors — with Shuaijun Liu, Chengyu Wu, and Feiyang You recurring — forming the basis of the collaboration score of 55.
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 are computed from a small curated snapshot: only two reviewed 2026 arXiv papers back the collaboration, focus, and reproducibility axes.
- Output, impact, and activity scores are missing — no publication counts, citation data, or multi-year publication cadence were available.
- No ORCID, institution, or country is on file; the profile is name-resolved from public bibliographic data, and 'Shuyang Hao' may not refer to a unique individual.
- Both cataloged papers are arXiv preprints (2026); no peer-reviewed venue record is included in the snapshot.