Contributor to open LLM post-training recipes and replayable forecasting-agent benchmarks; 2 reviewed 2026 papers
2
papers here
0
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-28) 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 publication count, venue record, or yearly output total is available in this snapshot, so the output axis is not scored. It would normally measure the volume and cadence of peer-reviewed and preprint publications over time.
Impact—
Citation counts, h-index, and downstream-usage signals are absent from the input data, so impact cannot be scored. This axis would measure how often and how widely the researcher's work is cited, adopted, or built upon by others.
Focus—
Only two reviewed papers underpin this profile, too thin to characterize thematic concentration, so the focus axis is unscored. It would measure how tightly the publication record clusters around a coherent set of research topics.
Collaboration67/100
Scores 67 out of 100, derived from 28 distinct coauthors recorded in this profile's bibliographic snapshot. Twelve recur frequently — among them Rongzhi Zhang, Chao Zhang, Liqin Ye, and Haorui Wang — indicating a stable core team rather than purely one-off collaborations. Both reviewed 2026 papers are multi-authored with Yuan He in a middle slot, consistent with active but non-lead participation in group projects.
Activity—
Recent-year publication velocity cannot be computed from a snapshot containing only two 2026 entries, so the activity axis is not scored. It would measure output frequency and engagement in the most recent publication years.
Reproducibility—
No author-level reproducibility score exists: artifact and replication data are too thin to compute it. The axis would measure whether published work ships reusable code, data, and documentation. Evolve does flag the two reviewed papers as medium and high repro-readiness per paper, but no author-level rating.
Research summary
This is an early-stage, name-resolved profile for Yuan He, built from public bibliographic data and two papers reviewed in the Evolve catalog. No institutional affiliation, country, or ORCID record was available, and the name is common enough that disambiguation rests on co-authorship overlap rather than verified biography. Both reviewed works appeared on arXiv in 2026 and sit in applied large language model engineering. Rufus-Air contributes to an open eight-stage post-training recipe built on GLM-4.5-Air-Base that Evolve judges to outperform the official GLM-4.5-Air release, though the review flags the stage-order rationale as untested and rates reproducibility readiness medium. Forecast-Dojo provides replayable environments for benchmarking and training LLM forecasting agents; Evolve rates its reproducibility readiness high and finds the core finding credible — the research tooling helps all twelve models tested, yet every model still trails the market. On both papers Yuan He is a middle author and not the corresponding author, so current evidence supports a contributor rather than project-lead positioning. The collaboration axis scores 67 out of 100 from 28 distinct coauthors, with a recurring core of twelve partners including Rongzhi Zhang, Chao Zhang, and Liqin Ye. Output, impact, focus, activity, and reproducibility composites remain unscored for lack of data.
Collaboration pattern
Yuan He works inside a fairly wide but stable collaboration network: 28 distinct coauthors in the snapshot, with a recurring core of twelve, including Rongzhi Zhang, Chao Zhang, Liqin Ye, Haorui Wang, Fardin Ahmed, and Sudheer Chava. On both reviewed 2026 papers the position is middle author and not corresponding, suggesting contributions within team-led projects rather than solo or senior-lead efforts. The collaboration axis scores 67 out of 100 on this basis.
Research keywords
Highlights
Open post-training recipe beats an official release
Replayable environments for LLM forecasting agents
Stable collaboration core behind an early-stage profile
The snapshot records 28 distinct coauthors and a collaboration score of 67/100, with twelve recurring partners including Rongzhi Zhang, Chao Zhang, Liqin Ye, and Haorui Wang. Both 2026 entries list Yuan He as a middle author, not corresponding.
Timeline
- 2026Rufus-Air appears on arXiv: an open eight-stage post-training recipe on GLM-4.5-Air-Base, reviewed by Evolve with medium repro-readiness.
- 2026Forecast-Dojo appears on arXiv: replayable environments for benchmarking and training LLM forecasting agents, with high repro-readiness.
- 2026Both 2026 entries enter the Evolve catalog with Yuan He as a middle, non-corresponding author on each.
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 caveat: Yuan He is a common name; this profile is name-resolved from public bibliographic data only, with no institution, country, or ORCID record to confirm disambiguation.
- Bibliometrics come from OpenAlex-derived snapshots; five of six radar axes (output, impact, focus, activity, reproducibility) lack the data to be scored.
- Only two reviewed papers, both 2026 arXiv preprints, underpin the qualitative assessment; no peer-reviewed venue record is included.
- Yuan He is a middle, non-corresponding author on both reviewed works, so paper-level verdicts reflect team output, not solely this researcher.