Privacy and security of LLM agents at Tencent: leakage frameworks, platform mediation, and trustworthy agent memory
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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—
Not scored: this axis measures publication volume and cumulative productivity. Verified bibliographic coverage is still too thin for a fair score, so it stays unscored until the 2026 arXiv work is fully indexed.
Impact—
Not scored: this axis measures citation influence such as h-index and i10-index. The record shows 0 on both, with 2026 preprints too recent for citations, so no fair impact estimate is possible.
Focus0/100
The radar scores focus at 0/100, but the raw signals read as concentrated rather than scattered: the OpenAlex field distribution splits into three near-identical labels — AI security and privacy for LLM agents, and computer security for LLM-agent systems privacy — each at one third, and topic counts cluster on privacy-preserving technologies (8) and adversarial robustness (6). The zero appears to reflect fragmented field labels in the taxonomy rather than genuinely dispersed research.
Collaboration54/100
Collaboration scores 54/100. OpenAlex records 14 distinct co-authors within the recent three-year window, and the Evolve catalog shows mid-list positions on CIPL and PAPC with a last-author placement on Semantic-TVM, though none of the three papers credits a corresponding-author role. Recurring names include Tao Huang, Guolong Zheng, and Guosen Wu, pointing to a stable but still-forming team pattern.
Activity67/100
Activity scores 67/100. OpenAlex attributes 9 works to this researcher in the recent three-year window, and the year-by-year count places all 9 in 2026, matching three reviewed arXiv preprints from the same year. The recent burst indicates an actively publishing profile, though the single-year concentration means a longer-term cadence cannot yet be judged.
Reproducibility55/100
Reproducibility scores 55/100 (reproMean) across 3 reviewed papers (reproPapers), all rated medium. CIPL reports results coherent across 5 providers but without statistical testing; PAPC's zero-leak results are partly by construction; Semantic-TVM reports 84% versus 52% task success on a small evaluation with no statistics and no code. Methodologically consistent work whose evaluation rigor and artifact release still lag.
Research summary
Chen Hou is a researcher at Tencent (China) whose published work centers on the security and privacy of large language model (LLM) agents. The OpenAlex snapshot attributes 9 works to this researcher within the recent three-year window — all landing in 2026 — across topics led by privacy-preserving technologies in data (8) and adversarial robustness in machine learning (6), with smaller clusters in privacy and data protection (3), access control and trust (2), and stochastic gradient optimization (2). Evolve has reviewed three 2026 arXiv preprints: CIPL, a channel-aware framework for recoverable privacy leakage in LLM agents whose results stay coherent across 5 providers; PAPC, an event-level platform-mediation design for privacy-propagation externalities in multi-agent workflows; and Semantic-TVM, a structure-preserving trustworthy virtual memory scheme whose span-level projection lifts masked-agent task success from 52% to 84%. Author positions — middle on the first two, last on Semantic-TVM — and 14 distinct recent co-authors (collaboration 54/100, activity 67/100) describe an actively contributing member of a forming group rather than a solo line. Reproducibility averages 55/100 across the three papers, all rated medium: the methods are coherent, but statistical testing, larger evaluations, and code release are missing. Citation impact is not yet measurable — h-index and i10-index stand at 0 — so this reads as an early-stage, output-oriented profile.
Collaboration pattern
OpenAlex records 14 distinct co-authors in the recent three-year window (collaboration score 54/100). Recurring names include Tao Huang, Guolong Zheng, and Guosen Wu, with Yu Li, Qikun Cai, Jiayang Meng, Xu Yang, Xuechao Yang, and Feng Xia also frequent. Across the three reviewed papers this researcher sits mid-list twice and once in the last position, never as corresponding author — a pattern consistent with contributing inside an established Tencent group rather than leading an independent line.
Research keywords
Highlights
Formalizing exposure versus recovery in agent privacy leakage
Event-level platform mediation for privacy propagation in AI workflows
Trustworthy virtual memory recovers masked-agent utility
Timeline
- 2026Publishes CIPL on arXiv: a channel-aware framework for recoverable privacy leakage in LLM agents, with the researcher in a middle author position.
- 2026Publishes PAPC on arXiv: event-level platform mediation for privacy-propagation externalities in AI-mediated multi-agent workflows, again as a middle author.
- 2026Publishes Semantic-TVM on arXiv: structure-preserving trustworthy virtual memory for memory-augmented and tool-using agents, in the last author position.
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.
- Bibliometric metrics (h-index, i10-index, topic and co-author counts, year counts) come from an OpenAlex snapshot; no other citation database is included.
- Output and impact dimensions are unscored: citation coverage for the 2026 arXiv preprints has not accumulated, and h-index and i10-index are 0 at snapshot time.
- All reviewed papers are 2026 arXiv preprints rated medium for reproducibility; no peer-reviewed venue record is available in this data.
- 'Chen Hou' is a common Chinese name and the co-author list contains duplicates and the researcher's own entry; this profile is name-resolved from public OpenAlex bibliographic data and may be incomplete.