UNC Chapel Hill ML researcher spanning clinical multimodal LLMs, efficient multimodal learning, and LLM unlearning
5
papers here
4
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
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—
Not yet scored: the underlying data on publication volume, paper counts, and venue throughput is insufficient in this profile. This axis would measure research output over time once fuller bibliographic records become available.
Impact—
Not yet scored: citation counts, h-index, and downstream uptake data are unavailable here. This axis would measure scholarly influence — how strongly the work is cited and adopted — once such bibliometric evidence is collected.
Focus0/100
The focus score of 0 reflects maximal breadth rather than absence of work: the field distribution splits evenly at 0.25 across four areas — clinical multimodal LLMs, multimodal AI systems, computational drug repurposing/biomedical NLP, and trustworthy ML (LLM unlearning). The four reviewed 2026 papers fall one per area, so no single field yet dominates the catalog.
Collaboration72/100
The collaboration score of 70 rests on 32 distinct coauthors across the reviewed work, with a recurring core of twelve names including Zijie Liu, Jinhao Duan, Inyoung Choi, Sukwon Yun, Jiayi Xin, Jie Peng, Qi Long, Pan Wang, Siwei Song, Hui Ji, Siqi Cao, and Heng Yu. Authorship positions vary — middle on the clinical and survey papers, last on the unlearning and drug-repurposing papers — with no corresponding-author role recorded.
Activity—
Not yet scored: this profile lacks a full publication history and per-year output, so recent publishing cadence cannot be computed. This axis would measure research momentum and throughput from longitudinal bibliographic records.
Reproducibility55/100
Reproducibility scores 55, a mean computed over the four reviewed papers, each rated medium reproducibility readiness. The review verdicts point the same way: M2G-LLM's ablations support its design, while DrugReason reports single-seed results without Macro-F1, and the tabular unlearning work offers small-scale single-run evidence — designs are documented but empirical backup remains thin.
Research summary
Tianlong Chen is a machine-learning researcher at the University of North Carolina at Chapel Hill whose reviewed work spans four equally weighted areas: machine learning for healthcare, multimodal AI systems, computational drug repurposing/biomedical NLP, and trustworthy ML centered on LLM unlearning. All four papers in the Evolve catalog are 2026 arXiv preprints, and the field distribution splits evenly at 0.25 each, signaling an active agenda whose center of gravity is still forming rather than a single dominant theme. The methodological through-line is combining large language models with structured or auxiliary evidence: M2G-LLM injects GNN-built multimodal context into a frozen LLM for clinical prediction on MIMIC; DrugReason routes knowledge-graph and language evidence to two 3B experts with mutual distillation for drug repurposing; a tabular unlearning paper proposes CAU, relaxing retain rows near the forget set; and a survey claims the first Model-Algorithm-System taxonomy for efficient multimodal learning. Chen has worked with 32 distinct coauthors, including a recurring core such as Qi Long, Jinhao Duan, and Sukwon Yun. Authorship sits middle on the clinical and survey papers and last on the unlearning and drug-repurposing papers, with no corresponding-author role recorded. Reproducibility readiness is rated medium across all four papers (mean 55), and review verdicts flag single-run or single-seed evidence as the main limitation.
Collaboration pattern
Chen's reviewed work shows a broad, team-based pattern: 32 distinct coauthors, with a stable core of twelve recurring names including Zijie Liu, Jinhao Duan, Inyoung Choi, Sukwon Yun, Jiayi Xin, Jie Peng, Qi Long, Pan Wang, Siwei Song, Hui Ji, Siqi Cao, and Heng Yu. Authorship ranges from middle positions on the clinical and survey papers to last positions on the unlearning and drug-repurposing papers; none lists him as corresponding author.
Research keywords
Highlights
M2G-LLM: graph-built multimodal context for clinical prediction
First Model-Algorithm-System taxonomy for efficient multimodal learning
CAU: easing forget-retain interference in tabular LLM unlearning
DrugReason: routing knowledge-graph and language evidence for drug repurposing
Timeline
- 2026Co-authored M2G-LLM, combining graph-built multimodal context with a frozen LLM for clinical prediction on MIMIC data (middle authorship).
- 2026Co-authored a comprehensive survey of efficient multimodal learning claiming the first Model-Algorithm-System taxonomy (middle authorship).
- 2026Last-position author on the CAU work mitigating forget-retain interference in tabular LLM unlearning.
- 2026Last-position author on DrugReason, multi-view knowledge-graph and language reasoning for drug repurposing.
Papers in the Evolve catalog
M2G-LLM: Enhancing Clinical Prediction via Multimodal Graph Reasoning and LLM Context Injection
From Models to Systems: A Comprehensive Survey of Efficient Multimodal Learning
When Retain Constraints Conflict: Mitigating Forget-Retain Interference in Tabular Data
DrugReason: Dynamic Multi-View Reasoning over Knowledge Graph and Language Evidence for Drug Repurposing
MASkills: Continual Skills Optimization for Multi-Agent LLM Systems
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 here cover only the four 2026 arXiv papers reviewed by Evolve; no citation counts, h-index, or venue acceptance data were provided, which is why output, impact, and activity are unscored.
- All cataloged papers are recent arXiv preprints; peer-review status and final publication venues are unknown.
- Radar scores are Evolve-computed bibliometric indicators (collaboration 70, focus 0, reproducibility 55) and should not be read as judgments of research quality.
- Profile name-resolved from public bibliographic data via the provided ORCID (0000-0001-7774-8197); country and detailed affiliation history were not available.