Xiamen University researcher spanning LLM alignment data, NLP privacy, and curvature-flow drug–target prediction
3
papers here
3
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—
Insufficient data to score. This axis measures publication volume and productivity over time; the current evidence base is only three 2026 arXiv preprints, so no output estimate is attempted.
Impact—
No citation or impact metrics are available yet. This axis would capture citations, downstream uptake, and influence of the work; with all three papers being fresh 2026 preprints, it is left unscored.
Focus0/100
The focus score is 0, driven by an even field distribution: the three cataloged papers split one third each across LLM alignment and human-in-the-loop annotation, NLP security and privacy, and AI-driven drug discovery. This reads as a broad methodological portfolio — annotation efficiency, inference-time privacy, and graph-curvature modeling — rather than a concentrated single-track agenda, though the sample is small.
Collaboration56/100
Collaboration scores 56 on the radar, grounded in 16 distinct coauthors across the three papers. Twelve frequent collaborators recur, including Lei Yang, Mengyin Liu, Jia Wang, and Hangyu Guo. Zhao appears in middle position on onPanda and last position on the other two papers, none as corresponding author, indicating a consistent contributor role within multi-author teams rather than solo work.
Activity—
Not enough temporal data to score. This axis measures publishing cadence and recent activity; only three papers, all dated 2026, are on record, so no activity trend can be established.
Reproducibility55/100
Reproducibility scores 55, a mean computed across three scored papers. Each carries a medium repro-readiness rating, consistent with Evolve's verdicts: onPanda's evidence base is small, CROSS-MAP reports single-run statistics, and CurvFlow-DTA's stats are untested despite protocol-matched Davis/KIBA gains. The mid-range score reflects solid protocols offset by unverified statistical rigor.
Research summary
Liang Zhao is a researcher at Xiamen University whose cataloged work spans three equally weighted areas — natural language processing for LLM alignment and human-in-the-loop annotation, NLP security and privacy, and computational biology for AI-driven drug discovery — each accounting for one third of the three Evolve-reviewed papers, all 2026 arXiv preprints. The first, onPanda, introduces token-level correction for annotating on-policy alignment data, cutting annotation time 52% versus post-editing, though the evidence base is small and the training benefit untested. The second, CROSS-MAP (LLMs as Linguistic Chameleons), decouples semantics from structure to sanitize prompts sent to private LLM APIs, with strong empirical gains but single-run statistics. The third, CurvFlow-DTA, applies dual-graph Forman-Ricci curvature flow with pair-conditioned horizons to drug–target affinity prediction, showing protocol-matched gains on Davis and KIBA. Across the set Zhao holds middle and last author positions without corresponding-author status, works with 16 distinct coauthors, and all three papers carry medium reproducibility readiness. The radar focus score of 0 mirrors this three-way split, while collaboration (56) and reproducibility (55) sit mid-range.
Collaboration pattern
The three cataloged papers involve 16 distinct coauthors, giving a collaboration score of 56. Twelve recur as frequent collaborators — among them Lei Yang, Mengyin Liu, Jia Wang, and Hangyu Guo, plus Zheng Ge, Kang An, Binxing Jiao, Qi Han, Daxin Jiang, Siqi Shen, Xiangyu Zhang, and Yuzhu Mao. Zhao's positions (middle on one paper, last on two, none corresponding) point to steady team-based work spanning all three research lines.
Research keywords
Highlights
Token-level correction speeds on-policy alignment annotation by 52%
CROSS-MAP decouples semantics from structure for private LLM API use
Dual-graph Forman-Ricci curvature flow for drug–target affinity
Timeline
- 2026Publishes CurvFlow-DTA, applying dual-graph discrete Ricci curvature flow to drug–target affinity prediction with protocol-matched gains on Davis and KIBA.
- 2026Publishes CROSS-MAP (LLMs as Linguistic Chameleons), decoupling semantics and structure for privacy-preserving LLM communication.
- 2026Publishes onPanda, a token-level correction method for annotating on-policy alignment data for LLMs and agents, cutting annotation time 52% versus post-editing.
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.
- Liang Zhao is a common name; this profile is name-resolved from public bibliographic data and the three arXiv records, with no ORCID link provided.
- Bibliometric scores are derived from OpenAlex-sourced data; the output, impact, and activity axes are unscored because the underlying data is insufficient.
- All three papers are 2026 arXiv preprints, not yet peer-reviewed publications, and no citation data is available.
- Coauthor affiliations and country are not provided, so cross-institution collaboration patterns cannot be verified.