Yuyang Hu
Beijing Academy of Artificial Intelligence · China
BAAI researcher building agentic LLM systems — skill distillation for autonomous ML research and human-AI collaboration
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
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—
No output metrics are available in the current snapshot, so this axis is left unscored. It would measure publication volume, authorship position, and venue quality across the researcher's record over time.
Impact—
Citation and influence data are insufficient for scoring this axis. It would capture citation counts, h-index and i10-index trajectories, and uptake of the work by the wider research community once evidence accumulates.
Focus0/100
The radar scores focus at 0, consistent with a field distribution split exactly 50/50 between 'AI agents and human-AI collaboration' and 'AI agents / autonomous ML research (AI4AI)' across the cataloged 2026 papers. Topic counts stay within one agent-and-engineering cluster — AI-based Problem Solving and Planning (3), Artificial Intelligence Applications (3), Software Engineering Research (2) — so the work is topically adjacent rather than scattered, but the even two-way split anchors the measured score at zero.
Collaboration85/100
Collaboration scores 85, grounded in 71 distinct coauthors recorded in the OpenAlex snapshot. The network shows 12 recurring collaborators — including Honglin Guo, Tao Gui, Guanting Dong, and Yanjiang Liu — and both cataloged 2026 papers are middle-authorship team efforts, a pattern consistent with large-group lab production common to agentic LLM research.
Activity47/100
Activity scores 47: the snapshot counts 4 works in the last three years, and the countsByYear data places all of them in 2026. This compressed, recent burst — two arXiv preprints in the Evolve catalog plus additional 2026 entries — indicates an early-stage, actively publishing profile whose sustained output rhythm is not yet established.
Reproducibility38/100
Reproducibility averages 38 across the 2 assessed papers. The spread is instructive: Repo-To-Skill reaches medium readiness after distilling 1,000 repos with reported gains on 4 benchmarks, while Atria Dawn is rated low owing to self-reported attribution numbers and cherry-picking risk in its benchmark table. Replication confidence therefore varies by paper.
Research summary
Yuyang Hu is a researcher at the Beijing Academy of Artificial Intelligence (BAAI) whose cataloged work centers on agentic large language models — systems that plan, use tools, and improve from experience. This is an early-stage profile built from a single OpenAlex-resolved snapshot: both reviewed papers are 2026 arXiv preprints on which Hu is a middle author, and the citation record (h-index 0, i10-index 0) has not yet accumulated. The two papers define the profile's axis. 'Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills' distills 1,000 ML repositories into more than 5,000 verified skills that lift a fixed GPT-5.5 Codex agent on 4 benchmarks; Evolve rates its reproducibility medium while flagging cost and contamination for scrutiny. 'Atria Dawn: The Dawn of Agentic Superintelligence' is a model release report whose human-AI collaboration study is judged its most interesting part, though its benchmark table and self-reported attribution numbers carry cherry-picking risk; reproducibility is rated low. Bibliometrically, the field distribution splits evenly (50/50) between AI agents and human-AI collaboration and AI agents / autonomous ML research (AI4AI), with topics clustering in problem solving and planning (3), AI applications (3), and software engineering research (2). Collaboration is the standout signal: 71 distinct coauthors and 12 recurring collaborators yield a radar collaboration score of 85, while activity (47), reproducibility (38), and focus (0) reflect a profile just beginning to take shape.
Collaboration pattern
The OpenAlex snapshot records 71 distinct coauthors, driving a collaboration score of 85 — the profile's strongest radar dimension. Twelve recurring collaborators include Honglin Guo, Tao Gui, Yicheng Chen, Guanting Dong, and Yanjiang Liu. Both cataloged 2026 papers list Hu as a middle author without corresponding status, indicating contributions within large, team-based agentic-LLM projects rather than solo-led efforts.
Research keywords
Highlights
Repo-To-Skill: 1,000 ML Repos Distilled Into 5k+ Verified Skills
Atria Dawn: A Model Release Report Led by Its Human-AI Collaboration Study
Timeline
- 2026Co-authored 'Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills' (arXiv), a middle-authorship contribution to autonomous ML research.
- 2026Co-authored 'Atria Dawn: The Dawn of Agentic Superintelligence' (arXiv), in the AI agents and human-AI collaboration field.
- 2026OpenAlex countsByYear data records 4 works in 2026 — all of the snapshot's recent output falls in this single year.
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.
- Identity is name-resolved from public OpenAlex bibliographic data; no ORCID is linked, so attribution to this specific Yuyang Hu is provisional.
- All bibliometric indicators come from a single OpenAlex snapshot; no other scholarly databases contribute to this profile.
- Only two 2026 arXiv preprints are in the reviewed catalog, and citation indicators stand at h-index 0 and i10-index 0, so impact cannot be meaningfully assessed yet.
- The reproducibility score averages just 2 assessed papers, and the output and impact radar dimensions are unscored due to insufficient data.