Eric Hanchen Jiang
University of California, Los Angeles · United States
Early-career UCLA researcher spanning multimodal AI safety, LLM multi-agent systems, and federated graph learning
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
1
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 scored yet: this OpenAlex record lacks verified venue and publication-type metadata, so yearly output by venue tier and author position cannot be computed. This axis would measure publication counts per year, venue quality, and the share of first or corresponding-author papers once a verified work list is linked.
Impact—
Not scored yet: citation signals are minimal at this career stage (h-index 1, i10-index 0) and no citing-work distribution is available, so influence cannot be estimated. This axis would measure total citations, citing venues, and independent uptake beyond self-citation as the record accumulates.
Focus0/100
The focus score of 0 reflects an evenly spread portfolio rather than a missing signal: the field distribution splits exactly into thirds — machine learning (AI safety, multimodal), LLM agent safety and federated graph learning, and machine learning (LLM multi-agent systems) — while the top OpenAlex topics (topic modeling 11, NLP techniques 9, adversarial robustness 8, planning 7, robotics RL 6) span five distinct areas.
Collaboration60/100
A collaboration score of 60 fits a mid-sized, team-based pattern: OpenAlex lists 19 distinct coauthors across 25 works in the last three years, with twelve recurring collaborators including Jinxi Yu, Levina Li, Dong Liu, Zhi Zhang, Wenxiao Zhao, Ying Nian Wu, Edward Sun, Yuchen Wu, Zixian Ma, Yijia Xiao, Xiaoyuan Yi, and Ranjay Krishna. All three reviewed papers are middle-author, non-corresponding contributions.
Activity95/100
Activity is the strongest axis at 95: indexed works climbed from 1 in 2023 to 4 in each of 2024 and 2025, then 17 in 2026 — a more than fourfold single-year jump — and OpenAlex credits 25 works over the last three years. All three Evolve-reviewed papers appeared as 2026 arXiv preprints, consistent with this peak year.
Reproducibility70/100
Reproducibility averages 70 across the 3 Evolve-reviewed papers. The federated graph learning paper (FGLGuard) is rated high readiness, while the VLM-safety benchmark paper (MPS-Bench/PRISM) and Codebook Agent are medium: Evolve flags label circularity and judge overlap in the first, and single-run evidence with a partly circular collapse premise in the second.
Research summary
Eric Hanchen Jiang is a researcher at the University of California, Los Angeles whose work spans AI safety and large language model systems. OpenAlex records 25 works over the last three years, with annual output rising from 1 in 2023 and 4 in each of 2024–2025 to 17 in 2026, alongside 19 distinct coauthors — a steep early-career ramp on a still-thin citation record (h-index 1, i10-index 0). Three 2026 arXiv papers reviewed by Evolve frame the current agenda: MPS-Bench and the PRISM deferral monitor (0.978 AUC) for personalized safety failures in vision-language models, where visual input overrides prior knowledge; Codebook Agent, which amortizes communication-topology design for LLM multi-agent systems with vector-quantized codebooks (beating GTD by 1.6 points at 2.4 ms); and FGLGuard, which trains multi-agent safety guards via federated graph learning without sharing raw data. OpenAlex topic signatures — topic modeling (11), NLP techniques (9), adversarial robustness (8), planning (7), robotics RL (6) — and an even three-way field split indicate deliberate breadth rather than a settled niche. This is an early-stage, name-resolved profile.
Collaboration pattern
Works in middle-author team roles on all three reviewed papers, none as corresponding author, within a wide network of 19 distinct coauthors over the last three years. Twelve recurring collaborators include Jinxi Yu, Levina Li, Dong Liu, Zhi Zhang, Wenxiao Zhao, Ying Nian Wu, Edward Sun, Yuchen Wu, Zixian Ma, Yijia Xiao, Xiaoyuan Yi, and Ranjay Krishna — a pattern of project-based teams rather than a single long-term partnership.
Research keywords
Highlights
MPS-Bench and PRISM: benchmarking personalized safety in vision-language models
Codebook Agent: amortized communication-topology design for LLM multi-agent systems
FGLGuard: privacy-preserving multi-agent safety via federated graph learning
Timeline
- 2023First recorded OpenAlex-indexed output: 1 indexed work.
- 2024Output grows to 4 indexed works.
- 2025Sustains 4 indexed works as the coauthor network broadens toward 19 distinct collaborators.
- 2026Peak year with 17 indexed works, including all three Evolve-reviewed arXiv papers (VLM personalized safety, Codebook Agent, and FGLGuard).
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.