Sun Yat-sen University researcher across vision-language agents, agent memory, and tabular foundation models
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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: OpenAlex returns no usable output metrics for this profile (h-index 0, i10-index 0, only 3 indexed works), so publication volume and consistency cannot be assessed yet. This axis would measure the scale and regularity of research output over time.
Impact—
Not scored: the snapshot contains no citation counts, so scholarly influence is unknown. This axis would measure citations, h-index trajectory, and uptake of the three 2026 preprints once indexing catches up with the record.
Focus0/100
Scores 0/100 because the work splits perfectly evenly across three fields: the OpenAlex field distribution assigns 33.3% each to computer vision (generative models and design automation), machine learning (tabular deep learning / foundation models), and natural language processing (conversational AI agents), and the five recorded topics each appear once. This indicates breadth rather than a concentrated line; no dominant theme emerges from the current sample.
Collaboration70/100
Scores 70/100, supported by 33 distinct coauthors in the OpenAlex record despite only 3 indexed recent works. Recurring collaborators on the reviewed papers include Junyan Ye, Dongzhi Jiang, Zichen Wen, Jun He and the group credit 'Xiaomi-TabLDM Team', and all three papers place Wei Liu in a middle-author position without corresponding authorship — a contributor role within larger team-based groups.
Activity40/100
Scores 40/100. OpenAlex counts-by-year show 3 works in 2026 with no earlier indexed years, and the last-3-year count is likewise 3. All catalogued output is arXiv preprints from a single recent year, so the publishing cadence is real but the indexed history is short; a longer record would be needed to judge sustained activity.
Reproducibility43/100
Scores 43/100, the mean across the 3 reviewed papers. Reproducibility readiness is rated low for 'Editable Visual Design' (showcase-only validation, no metrics) and medium for both 'RuleMem' (strong results, but the headline margin is versus a baseline average, not the best baseline Zep) and 'Xiaomi-TabLDM' (solid across 4 benchmarks, yet no ablations, overlapping Elo confidence intervals, and unquantified test-time scaling).
Research summary
Wei Liu is a researcher at Sun Yat-sen University, China. The Evolve-reviewed record spans three areas in near-equal measure: the OpenAlex field distribution assigns about 33% each to computer vision (generative models and design automation), machine learning (tabular deep learning / foundation models), and natural language processing (conversational AI agents). Three 2026 arXiv preprints are catalogued, all with middle-author positions and none corresponding. 'Editable Visual Design' builds a vision-language-model agent plus image-simulator pipeline that yields editable HTML/CSS designs, though validation rests on showcase cases without metrics. 'RuleMem' induces RPC-filtered Horn-clause rules for long-term conversational agent memory, reporting strong results whose headline margin is measured against a baseline average rather than the strongest baseline (Zep). 'Xiaomi-TabLDM' is a technical report on a synthetic-pretrained in-context-learning tabular foundation model, with solid evidence across four benchmarks but no ablations. OpenAlex records 33 distinct coauthors and 3 works over the last three years, all dated 2026. Radar scores put collaboration at 70/100, activity at 40/100, and reproducibility at 43/100 (mean of 3 papers), while focus is 0/100, consistent with the even three-way split. Output and impact are unscored — h-index and i10-index both read 0 — so no citation-based claim is made.
Collaboration pattern
OpenAlex records 33 distinct coauthors against only 3 indexed recent works, pointing to dense team-based collaboration. Recurring names include Junyan Ye, Dongzhi Jiang, Zichen Wen, HaoDong Li, Jun He and the group credit 'Xiaomi-TabLDM Team', the latter tying the tabular-model work to industry participation. Across all three reviewed papers Wei Liu holds a middle-author position without corresponding authorship, consistent with a contributing member of larger lab or industry teams rather than a sole lead.
Research keywords
Highlights
Editable Visual Design: agentic pipeline for editable HTML/CSS designs
RuleMem: Horn-clause rule memory for long-term conversational agents
Xiaomi-TabLDM: synthetic-pretrained tabular foundation model report
Timeline
- 2026Publishes 'Editable Visual Design', a VLM-agent and image-simulator pipeline producing editable HTML/CSS designs (arXiv preprint, middle-author position).
- 2026Publishes 'RuleMem', introducing active rule memory based on RPC-filtered Horn-clause rules for long-term conversational agents (arXiv preprint, middle-author position).
- 2026Co-authors the 'Xiaomi-TabLDM' technical report on a synthetic-pretrained tabular foundation model with in-context learning (arXiv preprint, middle-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.
- All bibliometrics come from OpenAlex only; h-index and i10-index are recorded as 0, reflecting sparse indexing of an early record rather than measured citation impact.
- 'Wei Liu' is a common name and no ORCID is linked; this profile is name-resolved from public OpenAlex bibliographic data and may mix identities.
- The evidence base is 3 arXiv preprints from 2026 reviewed by Evolve; peer-review status, venues, and citation counts are not yet available.
- Output and impact dimensions are unscored due to insufficient data; no publication counts beyond the 3 indexed works or any citation figures exist in the snapshot.