Peking University researcher across LLM agents, watermarking, mobility ML, and autonomous ferroelectrics discovery
4
papers here
4
reviewed by Evolve
2
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—
Not scored: no publication-count or venue-quality signals beyond the 6 recent works are available, and h-index and i10-index are recorded as 0. This axis would measure overall research output volume, publication venues, and career-level productivity.
Impact—
Not scored: citation counts, h-index-based influence, and downstream adoption are not yet available in the data (both indices read 0). This axis would measure citations, usage of released benchmarks or code, and influence on follow-up work.
Focus0/100
Score 0 reflects maximal dispersion rather than depth in one line: the OpenAlex field distribution splits evenly, 0.25 each, across LLM agent evaluation, text watermarking/NLP security, urban mobility analytics, and ferroelectric autonomous experimentation, while five listed topics carry one work each. The four reviewed 2026 papers sit in four different fields with no overlap, so no dominant research line emerges from the current record.
Collaboration66/100
Score 66 from 27 distinct co-authors in the recent window, a moderately broad network. The ferroelectrics line connects to a scanning-probe community — frequent co-authors include Sergei V. Kalinin, Lane W. Martin, Boris Slautin, and Zhongli Fang — while other papers bring separate collaborator sets such as Yiran Chen, Xiaoyan Sun, and Yi Wang, consistent with field-spanning, team-based work.
Activity57/100
Score 57 based on 6 works in the last three years, all dated 2026 in the record. Activity is current but shallow in visible history: the year-by-year counts list only 2026 (6 works), with no earlier years recorded, and four of these outputs are arXiv preprints reviewed by Evolve, indicating an active, early-visibility phase.
Reproducibility57/100
Score 57 as the mean repro-readiness across the 4 reviewed papers, which spread widely: the SPARC ferroelectrics microscopy study rates high, the DAREBench benchmark and the ODIN mobility work rate medium, and the ToSS watermarking paper rates low, with editing-robustness untested. The average signals adequate but uneven methodological transparency.
Research summary
Yu Liu is a researcher at Peking University whose profile, name-resolved from OpenAlex bibliographic data, is defined by unusual breadth. The record shows 6 works in the last three years, all dated 2026, and 27 distinct co-authors; Evolve has reviewed four 2026 arXiv preprints spanning four evenly weighted fields: LLM agent evaluation, NLP security (text watermarking), urban mobility analytics, and autonomous experimentation in ferroelectrics. The methodological throughline is machine learning applied to measurement and evaluation problems. As first author, Liu led DAREBench, a deployment-aware benchmark unifying 233 tasks and 35 models with evidence-audited scoring, and the SPARC human-agent microscopy study that discovered polarity-alternation control of PZT in-plane superdomains. On the urban mobility paper Liu is last author, contributing to ODIN, which infers origin-destination flows from aggregate counts across 12 datasets; on the food-safety watermarking paper, a middle author position. Reproducibility signals are mixed: a mean readiness of 57 across 4 reviewed papers, from high (ferroelectrics microscopy) to low (watermarking). Citation indicators are not yet available — h-index and i10-index are recorded as 0 — so output and impact cannot be scored, making this an early-stage profile whose substance rests on the reviewed 2026 work itself.
Collaboration pattern
With 27 distinct co-authors across a handful of 2026 papers, collaboration is wide and field-segmented. The ferroelectrics work clusters with scanning-probe researchers — Sergei V. Kalinin, Lane W. Martin, Boris Slautin, Zhongli Fang, Ching-Che Lin — while the ML papers list separate collaborators (Yiran Chen, Xiaoyan Sun, Yi Wang, Jun Dai, Ping Chen, Lingyun Zhang, Jaegyu Kim). Liu takes first author on two papers, last on one, middle on one, none corresponding.
Research keywords
Highlights
DAREBench: deployment-aware, reliable evaluation of models as agents
SPARC human-agent microscopy discovers ferroelectric superdomain control
ODIN infers origin-destination flows from aggregated mobility counts
ToSS: multi-bit watermarking for LLM-generated food safety content
Timeline
- 2026Bibliometric record shows 6 works in the last three years, all dated 2026, alongside 27 distinct co-authors
- 2026First-authored DAREBench appears on arXiv: a deployment-aware agent benchmark unifying 233 tasks and 35 models with evidence-audited scoring
- 2026Middle-author ToSS framework adds multi-bit traceability to LLM watermarking for food-safety content
- 2026First-authored SPARC human-agent microscopy study reports polarity-alternation control of PZT in-plane superdomains
- 2026Last-author ODIN paper infers origin-destination interactions from aggregated mobility counts across 12 datasets
Papers in the Evolve catalog
Inferring Urban Mobility Interactions from Aggregated Dynamics
- first author
Human-agent discovery of reconfigurable in-plane ferroelectric superdomain control
A Trustworthy Watermarking Framework for LLM-Generated Food Safety Content
- first author
DAREBench: Deployment-Aware and Reliable Evaluation of Models as Agents
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 indicators are computed from OpenAlex only; h-index and i10-index are recorded as 0, so the output and impact dimensions are unscored rather than measured as zero.
- "Yu Liu" is a very common name; this profile is name-resolved from public bibliographic data (OpenAlex) with no ORCID, and the evenly split multi-field record may partly reflect same-name conflation.
- All four reviewed papers are 2026 arXiv preprints; citation and venue information has not accumulated, and Evolve verdicts are expert reviews, not peer-reviewed assessments.
- Reproducibility scoring covers only the 4 Evolve-reviewed papers (mean 57); topic and year counts come from the same bibliometric snapshot.