Researcher at Minjiang University working on LLM-agent privacy leakage and lightweight time-series forecasting modules
3
papers here
3
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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 score yet: this axis would measure publication volume and authorship roles across venues, but the current evidence is limited to three 2026 arXiv preprints, too little to characterize overall output.
Impact—
Impact is unscored: this axis would track citations and downstream uptake of the work, yet all cataloged papers are 2026 preprints and the snapshot contains no citation data, so no estimate is offered.
Focus0/100
Focus scores 0 because the field distribution splits exactly into thirds: 33% machine learning (time-series forecasting), 33% AI security & privacy (LLM agents) and 33% computer security (LLM-agent systems privacy). The two security strands are thematically adjacent, both on LLM-agent privacy, while the forecasting paper sits apart, so the profile reads as two parallel threads rather than one concentrated line.
Collaboration53/100
Collaboration scores 48, drawn from 10 distinct coauthors in the snapshot. The frequent-coauthor list repeats Chen Hou and Guosen Wu, marking the closest working ties, while Guolong Zheng, Hongkai Zhuang, Jiayang Meng, Xu Yang, Xuechao Yang and Feng Xia each appear once. The moderate score and modest pool fit an early-stage network of small arXiv teams rather than large consortiums.
Activity—
Activity is unscored: this axis measures publication cadence and recent output rate, but the snapshot spans a single year (2026) with three papers, which is insufficient to establish an output trajectory.
Reproducibility55/100
Reproducibility averages 55 across the three reviewed papers, each rated medium readiness. The pattern behind the score is consistent: Evolve describes coherent, honestly framed experiments — five providers for CIPL, three Transformers for the gate — but flags the same gap in all three, the absence of statistical significance testing, and notes that PAPC's zero-leak results are partly by construction. Replication looks feasible; confirmation awaits stronger evaluation.
Research summary
Tao Huang is a researcher at Minjiang University whose cataloged work sits at the intersection of LLM-agent security and applied machine learning. The Evolve catalog lists three 2026 arXiv preprints. Two are first-authored security papers: CIPL, a channel-aware framework formalizing exposure-versus-recovery for black-box privacy leakage in LLM agents — coherent across five providers though untested statistically — and PAPC, which proposes event-level platform mediation for privacy-propagation externalities in multi-agent workflows, with a sound concept but zero-leak results that are partly by construction. The third, a middle authorship, is a lightweight plug-in gate: a two-layer MLP reweighting covariates before the encoder in three Transformer forecasters, reporting mostly parity with honest framing. The field distribution splits evenly in thirds across machine learning (time-series forecasting), AI security & privacy (LLM agents), and computer security (LLM-agent systems privacy), which is why the focus score is 0. Ten distinct coauthors appear in the snapshot, giving a collaboration score of 48; reproducibility averages 55 across the three papers, all rated medium readiness, with missing significance testing the recurring gap.
Collaboration pattern
The snapshot records 10 distinct coauthors and a collaboration score of 48. Chen Hou and Guosen Wu are the only names repeated in the frequent-coauthor list, marking them as the closest collaborators; Guolong Zheng, Hongkai Zhuang, Jiayang Meng, Xu Yang, Xuechao Yang and Feng Xia appear once each. Teams are small arXiv-scale groups, and the two first-authored security papers suggest Tao Huang drives the LLM-agent privacy line while collaborating on the forecasting thread.
Research keywords
Highlights
Channel-aware framework formalizes recoverable privacy leakage in LLM agents
Event-level platform mediation for privacy propagation in multi-agent workflows
Lightweight MLP gate reweights covariates in Transformer forecasters
Timeline
- 2026First-authors CIPL, a channel-aware framework for recoverable privacy leakage in LLM agents (arXiv preprint).
- 2026First-authors PAPC on platform mediation for privacy-propagation externalities in AI-mediated workflows (arXiv preprint).
- 2026Co-authors a lightweight plug-in gate for Transformer-based time-series forecasters (arXiv preprint).
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 come from the provided snapshot only; the output, impact and activity axes lack sufficient data and are left unscored.
- All three catalog entries are 2026 arXiv preprints; no peer-reviewed venue, citation count, h-index or ORCID data is available.
- 'Tao Huang' is a common name; this profile is name-resolved from public bibliographic data and may merge or split same-name authors.
- Reproducibility reflects Evolve's review of three papers only; no code-release or artifact availability data was provided.