Huanrui Yang

University of Arizona · United States

Unclaimed · AI-structured profileORCID OpenAlex

University of Arizona researcher in efficient multimodal AI systems and LLM unlearning with model compression.

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

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 yet scored. This axis would measure publication volume, venue spread, and yearly output from indexed records. Raw year-by-year counts exist in the snapshot, but the composite output score has not been computed, so no rating is offered here.

Impact—

Not yet scored. This axis would capture citation-based influence such as citation counts and per-paper reach. The snapshot reports h-index 22 and i10-index 30, but the composite impact score is not computed, so no rating is offered here.

Focus0/100

Scores 0, the lowest end of the focus axis, reflecting a broad portfolio. The field distribution splits evenly (50% multimodal AI systems, 50% LLM unlearning and model compression), and indexed topics span five clusters: Advanced Neural Network Applications (73), Domain Adaptation and Few-Shot Learning (35), Adversarial Robustness (24), Topic Modeling (17), and ML Data Classification (15). The record pairs two research lines rather than one narrow specialty.

Collaboration64/100

Scores 64. The snapshot records 24 distinct coauthors and a recurring core group, including Jingtong Hu, Yingyan Celine Lin, Beidi Chen, Mohit Bansal, Zhijian Liu, and Xiaoming Liu. Both cataloged 2026 papers list the researcher in a middle author position, consistent with contributing within larger multi-author teams rather than solo or corresponding-author work.

Activity100/100

Scores 100, the maximum on the activity axis. The snapshot indexes 66 works over the most recent three-year window, and yearly counts climb from 9 (2018) and 5 (2022) to 13 (2024) and 44 (2026), the largest single-year figure in the record. This indicates a sustained and currently accelerating publishing tempo across both research lines.

Reproducibility55/100

Scores 55, computed from Evolve's review of 2 papers, both rated medium reproducibility readiness. The unlearning paper was flagged for lacking significance tests and general-LLM evidence, while the survey offers a coherent taxonomy but no quantitative synthesis. Methods appear described well enough to attempt, though statistical rigor gaps remain.

Research summary

Huanrui Yang is a machine-learning researcher at the University of Arizona whose indexed record, resolved from OpenAlex, splits evenly between two lines: multimodal AI systems (50%) and LLM unlearning with model compression (50%). Reported bibliometrics include an h-index of 22 and an i10-index of 30. Activity is the strongest scored axis at 100: 66 works fall in the most recent three-year window, and yearly counts rise from single digits (9 in 2018, 5 in 2022) to 13 in 2024 and 44 in 2026, the largest single-year figure. Focus scores 0, reflecting a broad topic spread across five clusters led by Advanced Neural Network Applications (73), Domain Adaptation and Few-Shot Learning (35), and Adversarial Robustness (24). Collaboration scores 64, with 24 distinct coauthors and a recurring core that includes Jingtong Hu, Yingyan Celine Lin, Beidi Chen, Mohit Bansal, Zhijian Liu, and Xiaoming Liu; both 2026 Evolve-catalog papers list a middle author position. Reproducibility scores 55 from Evolve's review of two arXiv papers: a survey proposing a claimed-first Model-Algorithm-System taxonomy for efficient multimodal learning, and a study attributing unlearning reversal under 4-bit quantization to sharp forget basins — both rated medium readiness, with noted gaps in quantitative synthesis and significance testing. Output and impact axis scores remain uncomputed.

Collaboration pattern

The snapshot lists 24 distinct coauthors, with a stable repeat-collaboration core spanning Jingtong Hu, Pan Wang, Siwei Song, Yingyan Celine Lin, Zhijian Liu, Beidi Chen, Mohit Bansal, and Xiaoming Liu. Both 2026 catalog papers place the researcher in a middle author position without corresponding authorship, suggesting contribution within larger team-based projects rather than solo-led efforts. The pattern fits a collaborator bridging efficient-systems and unlearning lines.

Research keywords

Efficient multimodal learningMultimodal large language modelsModel compressionQuantizationLLM unlearningKV cache managementMixture-of-expertsDomain adaptation and few-shot learningAdversarial robustnessNeural network applicationsLoss landscapesTopic modeling

Highlights

First Model-Algorithm-System taxonomy for efficient multimodal learning
A 2026 arXiv survey organizes efficient multimodal learning — multimodal LLMs, model compression, KV cache management, and mixture-of-experts — into a claimed-first Model-Algorithm-System taxonomy; Evolve finds it coherent and well-sourced, though without quantitative synthesis.
Diagnosing unlearning reversal under quantization
A 2026 study traces unlearning reversal under 4-bit quantization to sharp forget basins in the retain-forget loss landscape interaction; Evolve judges the diagnosis plausible but notes missing significance tests and general-LLM evidence.
2026 publication surge across two research lines
Indexed works reach 44 in 2026, up from 13 in 2024, with 66 works across the most recent three-year window — the activity axis scores 100 as output spans both cataloged research lines.

Timeline

2018Earliest year in the indexed record, with 9 works counted in the snapshot's year-by-year counts.
2022Lowest point of the indexed record: 5 works, within a run of single-digit years (6 in 2019 and 6 in 2020).
2024Output rises to 13 indexed works, the highest annual total before 2026.
2026Both Evolve-cataloged arXiv papers appear: the efficient multimodal learning survey and the quantization-robust unlearning study; indexed works reach 44, the record's largest single-year count.

Papers in the Evolve catalog

Frequent co-authors

Jingtong Hu×2 Pan Wang×1 Siwei Song×1 Hui Ji×1 Siqi Cao×1 Heng Yu×1 Zhijian Liu×1 Huanrui Yang×1 Yingyan Celine Lin×1 Beidi Chen×1 Mohit Bansal×1 Xiaoming Liu×1

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 scores are derived from OpenAlex-indexed records (identity: openalex); the output and impact axis scores are not computed in this snapshot.
h-index (22) and i10-index (30) are quoted as reported; per-paper citation counts are not available in the input for verification.
Reproducibility (55) rests on only 2 Evolve-reviewed papers, both from arXiv and both in middle author positions, so the readiness estimate has a narrow base.
The frequent-coauthor list includes a self entry (Huanrui Yang), indicating name-disambiguation noise; this profile is name-resolved from public bibliographic data.