IR researcher at CUHK-Shenzhen: compact hash retrieval, private RAG, and web retrieval infrastructure for LLM agents
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
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—
Output is not scored: this snapshot lacks verified publication counts, venue lists, and total output volume. Once complete bibliographic records are linked, this axis would measure productivity volume and its distribution across venues and years.
Impact—
Impact is not scored: no citation counts, h-index, or downstream adoption data are available here. This axis would measure how strongly the reviewed preprints and future publications are cited and taken up by other work.
Focus0/100
The focus dimension scores 0 because the reviewed output splits evenly across three fields: information retrieval, privacy-preserving information retrieval (applied cryptography and IR systems), and computer networks / web retrieval infrastructure, each at 0.33 of the field distribution. By the metric's definition this registers as maximal breadth rather than concentration, though substantively it reflects a connected program spanning compact hashing, cryptographic retrieval, and edge delivery for LLM agents.
Collaboration14/100
Collaboration scores 14/100, based on only 1 distinct coauthor in the snapshot, with Peichun Hua the recurring frequent coauthor. All three catalogued 2026 preprints place Xiao in the last author position, though none designates a corresponding author. This points to a small, stable, possibly supervisory-led team pattern; larger or cross-institution collaboration is not visible in the current data and would raise this score as it appears.
Activity—
Activity is not scored: per-year publication rates and submission cadence cannot be computed from three records all dated 2026. This axis would measure publishing tempo and recent-year momentum once a fuller publication history is linked.
Reproducibility55/100
Reproducibility scores 55/100, the mean readiness across 3 reviewed papers, each independently rated medium. Evolve's verdicts explain the mid-range value: Spruce reports strong 0.2–3 s latencies but with NDCG-retention and leakage caveats; the Matryoshka hashing study is broad and well-ablated yet single-run with no code release; SemDN's motivating measurements are solid while its core coverage mechanism remains untested. Solid empirical grounding, with openness gaps.
Research summary
Yunming Xiao is a researcher at the Chinese University of Hong Kong, Shenzhen, whose reviewed work concentrates on making dense retrieval compact, fast, and private for LLM-era applications. The Evolve catalog includes three 2026 arXiv preprints, in all of which Xiao holds the last author position. Spruce addresses privacy-preserving retrieval-augmented generation under two-server secure multi-party computation, using 128-bit deep hashes to cut private retrieval to 0.2–3 seconds per query. Matryoshka Hash Representations introduces a two-stage nested binary hashing scheme for model-aware compact semantic retrieval that outperforms compact baselines across all tested budgets. SemDN proposes a content-delivery-network-style edge infrastructure that ships semantic chunks to LLM agents, backed by motivating measurements of web retrieval behavior. Together the three papers span three fields in equal measure — information retrieval, privacy-preserving IR at the intersection of applied cryptography and IR systems, and computer networks / web retrieval infrastructure — giving the profile a breadth-oriented focus score of 0. Collaboration is narrow in this snapshot: one distinct coauthor (Peichun Hua) appears, and the collaboration dimension scores 14. Reproducibility is moderate, with a mean readiness score of 55 across all three reviewed papers, each rated medium; Evolve's verdicts flag untested mechanisms, single-run results, and missing code releases as caveats. Output, impact, and activity metrics are not yet available in the underlying data.
Collaboration pattern
The snapshot records a single distinct coauthor, Peichun Hua, as the recurring frequent collaborator, yielding a low collaboration score of 14/100. All three reviewed 2026 preprints place Xiao in the last author position, with none designating a corresponding author — a pattern consistent with a senior or supervisory role on small, stable teams rather than large multi-institution collaborations. Broader co-authorship networks cannot be assessed from the current data.
Research keywords
Highlights
Spruce: private RAG retrieval in 0.2–3 seconds per query
Matryoshka hashing for model-aware compact semantic retrieval
SemDN: rethinking web retrieval infrastructure for LLM agents
Timeline
- 2026Publishes Spruce: scalable private outsourced retrieval using compact embeddings, reaching 0.2–3 s per query under two-server MPC.
- 2026Publishes Matryoshka Hash Representations, a two-stage nested binary hashing scheme for model-aware compact semantic retrieval.
- 2026Publishes SemDN, a CDN-like semantics delivery network that rethinks web retrieval infrastructure for LLM agents.
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 dimensions derive from OpenAlex-sourced data; the output, impact, and activity axes lack sufficient data and are left unscored.
- The evidence base is three 2026 arXiv preprints reviewed by Evolve; peer-reviewed venue acceptance and citation history are not reflected.
- All three catalogued papers place the researcher in the last author position; finer contribution roles beyond author order are unavailable.
- This profile is name-resolved from public bibliographic data linked to ORCID 0000-0002-4913-4881; 'Yunming Xiao' may correspond to multiple individuals.