Tencent researcher in NLP and multimodal AI: video-generation benchmarks, legal-reasoning LLMs, omni-modal robustness
3
papers here
0
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
3
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) 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 OpenAlex record lacks the normalized publication-volume and venue-quality detail needed to rate productivity. This axis would measure annual work counts, publication venues, and authorship-role distribution relative to field norms.
Impact—
Impact cannot be scored: the record shows h-index 3 and i10-index 0, too little accumulated citation evidence for a meaningful rating. This axis would measure citation counts, citation velocity, and uptake of Xuan's benchmarks by subsequent work.
Focus—
Focus is unscored: thematic classification across the full work set is not complete enough to rate concentration. The axis would measure how tightly output clusters around core topics such as topic modeling and multimodal learning versus dispersal across areas.
Collaboration72/100
The collaboration axis scores 72. The OpenAlex record lists 36 distinct coauthors, and twelve recur frequently — including Zhaolu Kang, Zhichao Hu, Yantao Liu, Tailong Luo, Leqi Zheng, and Eric Hanchen Jiang. Across the three cataloged 2026 papers Xuan appears in middle or last authorship rather than first, a pattern consistent with coordinating or supervising multi-author projects at Tencent rather than solo output.
Activity99/100
Activity scores 99, the strongest axis of this profile. Counts by year show 2 works in 2016, single works in 2018, 2019 and 2021, then a steep ramp: 4 works in 2024, 12 in 2025, and 13 already in 2026 — 29 works within the last three years. This near-continuous recent cadence indicates a researcher currently operating at full production capacity.
Reproducibility—
No reproducibility score yet: only one catalog paper (DramaChain Bench) carries a reproducibility assessment, rated low readiness. This axis would aggregate artifact availability, code release, and dataset openness across the full publication record.
Research summary
Richeng Xuan is an applied artificial-intelligence researcher at Tencent (China) whose OpenAlex-resolved record shows a young but sharply accelerating output: 34 indexed works in total, of which 29 fall in the last three years, including 12 in 2025 and 13 already in 2026. Citation impact has not yet accumulated — the h-index is 3 and the i10-index is 0 — marking this as an early-stage profile. Topic coverage is led by Topic Modeling (22 works), followed by Multimodal Machine Learning Applications (16), Natural Language Processing Techniques (14), Advanced Neural Network Applications (6), and GAN-based image synthesis (5), sketching a researcher who builds language- and multimodal-model systems and increasingly constructs the benchmarks that evaluate them. Evolve's catalog includes three 2026 arXiv papers. On DramaChain Bench, an end-to-end short-drama generation benchmark, Xuan is the last author; Evolve's review credits it as the first full-chain benchmark of its kind, with strong annotation and a candid agentic-LLM-judge audit, while noting items come from a proprietary pipeline and rankings rest on few models (reproducibility readiness: low). Two further catalog entries — JPO on structured legal reasoning for criminal judgment prediction and Modality Fault Lines on fragile omni-modal reasoning under corruption — extend the range from legal NLP to multimodal robustness.
Collaboration pattern
With 36 distinct coauthors and a collaboration score of 72, Xuan operates in team-based settings. Twelve collaborators recur, including Zhaolu Kang, Zhichao Hu, Yantao Liu, Tailong Luo, Leqi Zheng, and Eric Hanchen Jiang. The three cataloged 2026 papers place Xuan in middle or last authorship rather than first, consistent with a coordinating or supervisory role on multi-author arXiv projects at Tencent rather than solo work.
Research keywords
Highlights
DramaChain Bench: first end-to-end benchmark for short-drama generation
JPO: Juris Policy Optimization for structured legal reasoning in criminal judgment prediction
A 2026 arXiv paper in Evolve's catalog with Xuan as a middle author, introducing Juris Policy Optimization for structured legal reasoning in criminal judgment prediction. No Evolve review has been published for this entry yet, so no verdict, keywords, or reproducibility rating are available beyond the title.
Modality Fault Lines: structural corruptions expose fragile omni-modal reasoning
A 2026 arXiv catalog entry with Xuan as a middle author, examining how structural corruptions reveal weaknesses in omni-modal reasoning systems. Evolve has not yet reviewed this paper, so no verdict, keywords, or reproducibility assessment are available.
Timeline
- 2016Earliest indexed activity in the OpenAlex record: 2 works attributed to this author profile.
- 2021A single indexed work — the last before a multi-year gap in the record.
- 2024Output resumes at scale: 4 indexed works, the start of the current ramp.
- 202512 indexed works in a single year, the annual peak at that point.
- 202613 indexed works so far, including all three Evolve-cataloged arXiv papers: DramaChain Bench (last author), JPO, and Modality Fault Lines.
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 and may omit works not indexed there; the profile is name-resolved from public bibliographic data.
- Output, impact, focus, and reproducibility dimensions are unscored because the underlying data is insufficient; only collaboration (72) and activity (99) have computed scores.
- Only one cataloged paper (DramaChain Bench) carries an Evolve review, keywords, and a reproducibility rating; the other two entries are title-level only.
- Citation metrics (h-index 3, i10-index 0) reflect a young record and should not be read as a judgment of research maturity.