Brown University researcher spanning LLM agents, evaluation statistics, and trustworthy-AI calibration
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-08) 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—
Output and impact scores are not yet computed: this OpenAlex snapshot shows an h-index and i10-index of 0, consistent with all cataloged works appearing in 2026. The output axis would measure publication volume, venues, and career-level productivity.
Impact—
Impact cannot be scored yet: citation-backed aggregates are unavailable in this snapshot, with h-index and i10-index both registering 0 for works dated 2026. This axis would measure citation influence, uptake, and downstream reuse of the research.
Focus0/100
The focus score is 0, the profile's most dispersed reading: OpenAlex splits the field distribution evenly three ways (LLM-based agents, statistical evaluation, trustworthy-AI calibration, each 0.333), and topic assignments spread across five areas — software engineering research, testing and debugging, AI-based planning, adversarial robustness, and topic modeling (2/2/2/2/1). The breadth is real, but cohesion across the three threads is still forming.
Collaboration53/100
The collaboration score is 53, mid-range. OpenAlex records 13 distinct coauthors in the recent window, and the recurring collaborator list includes Yezhou Cheng, Runjia Du, Zeming Liu, Yilan Wei, Jingtao Zhang, and Yuan Xie. All three cataloged papers list Lyu in a middle author position and never as corresponding author, consistent with a contributor role inside small teams rather than an independent group lead so far.
Activity57/100
The activity score is 57. OpenAlex counts 6 works in the last three years, all indexed to 2026, and the Evolve catalog adds three 2026 arXiv preprints spanning agents, evaluation statistics, and calibration. Output is recent and concentrated in a single year, so a sustained multi-year publishing cadence cannot yet be verified from this snapshot.
Reproducibility85/100
Reproducibility is the profile's strongest axis at 85, computed as a mean over three assessed papers. Evolve rates two of them at high readiness — the hard-budget evaluation-limits paper and LWCal — and the agent-memory paper at medium, reflecting paired evidence drawn from a single templated benchmark and a single model.
Research summary
Hang Lyu is a researcher at Brown University whose recent work sits at the intersection of LLM-based agents, the statistics of machine-learning evaluation, and trustworthy-AI calibration. This profile is name-resolved from OpenAlex bibliographic data; the Evolve catalog reviews three 2026 arXiv papers, in each of which Lyu is a middle-listed, non-corresponding author. The first proposes scoped retrieval for persistent agent memory to prevent cross-family interference; the second proves sharp replication limits for hard-budget repeated LLM evaluation with matching designs and equal-budget evidence; the third introduces LWCal, a loss-weighted approach to isotonic calibration of tabular classifiers under noisy calibration labels. OpenAlex records 6 works over the last three years, all indexed to 2026, with 13 distinct coauthors. Topic assignments span software engineering research, software testing and debugging, AI-based problem solving and planning, adversarial robustness, and topic modeling, and the field distribution splits evenly across the three cataloged areas. A reproducibility mean of 85 across three assessed papers — two rated high and one medium readiness — is the profile's strongest signal, alongside honest, scope-aware verdicts on each paper.
Collaboration pattern
Collaboration is team-based: 13 distinct coauthors appear in the recent OpenAlex window, with recurring names including Yezhou Cheng, Runjia Du, Zeming Liu, Yilan Wei, Jingtao Zhang, and Yuan Xie. Across all three cataloged 2026 papers, Lyu sits in the middle of the author list and is never the corresponding author, suggesting a contributing-role pattern within small working groups rather than senior-led or solo projects. The mid-range collaboration score of 53 matches this picture.
Research keywords
Highlights
Scoped retrieval to protect agent memory
Sharp limits for honest uncertainty in LLM evaluation
LWCal: calibration under noisy labels
Timeline
- 2026arXiv preprint on scoped retrieval for persistent agent memory, preventing cross-family interference.
- 2026arXiv preprint proving sharp limits for honest uncertainty in hard-budget repeated evaluation.
- 2026arXiv preprint introducing LWCal, loss-weighted calibration for tabular classifiers with noisy calibration labels.
- 2026OpenAlex indexes 6 works to this year, with 13 distinct coauthors recorded across the recent window.
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 are from OpenAlex only (identity: openalex); no ORCID is linked, so this name-resolved profile may not capture every work by researchers named Hang Lyu.
- Citation-based output and impact scores are missing; h-index and i10-index read 0 in this snapshot, consistent with all indexed works dated 2026.
- Only three papers carry full Evolve reviews; verdicts and reproducibility ratings cover those papers alone and not the full output record.
- Author-position data shows middle, non-corresponding roles on all cataloged papers; contribution details beyond position are not available.