Yi Chen

Hong Kong University of Science and Technology · Hong Kong SAR China

Unclaimed · AI-structured profileOpenAlex

HKUST researcher pairing decision-reliability theory for LLM agents with multi-agent generative video systems

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-08) 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—

Output volume cannot be scored yet: the snapshot shows h-index 0, i10-index 0, and per-year counts covering only 2026 (3 works). This axis would measure cumulative publication volume and productivity trajectory once a longer record accumulates.

Impact—

Impact cannot be scored yet: the record carries no citation evidence beyond h-index 0 and i10-index 0. This axis would measure citations, citation velocity, and uptake of the 2026 papers in follow-up work.

Focus0/100

The focus score of 0 reflects low topical concentration rather than missing data. The field distribution splits exactly evenly — 0.5 machine learning (reinforcement learning and LLM agents) versus 0.5 multimedia computing and generative AI — and the topic list spans five clusters, led by adversarial robustness (2 papers) alongside human-technology interaction, AI-based problem solving and planning, and domain adaptation/few-shot learning. The two reviewed papers likewise sit in different fields.

Collaboration53/100

The collaboration score of 53 is computed from a recent-window network of 13 distinct co-authors. The frequent-coauthor list centers on Huatai Zhu, Qiang Chen, Ziqian Kou, Wenhao Li, Fei Wang, Yichao Cao, Xiu Su, Yuntian Xiao, Shoulong Zhang, Wenfeng Song, and Yan Wang. Both reviewed papers are multi-author efforts in which Chen holds the last position on one and a middle position on the other — a team-contributor pattern rather than a solo or corresponding-author profile.

Activity40/100

The activity score of 40 reflects a very recent record. The snapshot lists 3 works in the last three years, and countsByYear contains a single entry — 3 publications in 2026 — with nothing indexed before that year. Both catalog papers are 2026 arXiv preprints, and both h-index and i10-index are 0. The profile sits at the start of its publication trajectory, so activity should be re-read as future years accumulate.

Reproducibility38/100

Reproducibility scores 38 (mean 38 across the 2 papers with Evolve reproducibility reviews). Dual-Frontier is rated medium readiness: its certified verify-then-promote gates and strong benchmarks aid replication, but risk calibration is unclear. CogenPVG is rated low: the human study reports κ=0.33 with no significance tests, and no code or protocol availability is recorded in the review. The mean is coarse, resting on only two assessments.

Research summary

Yi Chen is an early-stage researcher at Hong Kong University of Science and Technology whose OpenAlex-resolved record spans two equally weighted areas — machine learning (reinforcement learning and LLM agents) and multimedia computing and generative AI, each at 0.5 in the field distribution. The profile is genuinely emerging: the snapshot records an h-index and i10-index of 0, three works in the last three years, and per-year counts covering only 2026 (3 publications). Two of these are in the Evolve catalog. On "Dual-Frontier: When Can an Agent Trust Its World Model?" (2026, arXiv), Chen is last author; Evolve's review credits the paper with strong theory — showing passive failure attribution is non-identifiable and motivating certified verify-then-promote gates built on conformal prediction and simultaneous confidence sequences — while flagging unclear risk calibration. On CogenPVG (2026, arXiv), Chen is a middle author; the work is described as the first ELM-guided multi-agent framework for persuasive video generation, though its human-study evidence is weak (κ=0.33, no significance tests). Topic clusters — adversarial robustness (2 papers), domain adaptation and few-shot learning, AI-based planning, and human-technology interaction — suggest a reliability-oriented agenda running across both agent and generative systems.

Collaboration pattern

Collaboration is team-based: the recent window counts 13 distinct co-authors, and the frequent list is anchored by Huatai Zhu, Qiang Chen, Ziqian Kou, Wenhao Li, Fei Wang, Yichao Cao, Xiu Su, Yuntian Xiao, Shoulong Zhang, Wenfeng Song, and Yan Wang. Across the two reviewed papers Chen appears as last author on Dual-Frontier and a middle author on CogenPVG, neither first nor corresponding — consistent with contributing within larger multi-author groups spanning both LLM-agent and generative-AI work.

Research keywords

world modelsLLM agentsfailure attributiondecision reliabilityconformal predictionmulti-agent LLM frameworkspersuasive video generationElaboration Likelihood Modeladversarial robustnessdomain adaptation and few-shot learning

Highlights

Theory-first result: passive failure attribution in agent world models is non-identifiable
As last author on Dual-Frontier (arXiv 2026), Chen contributes to a proof that passive failure attribution is non-identifiable, motivating certified verify-then-promote gates built on conformal prediction and simultaneous confidence sequences. Evolve rates the theory strong and the benchmarks strong, while flagging that risk calibration remains unclear; reproducibility readiness is medium.
First ELM-guided multi-agent framework for persuasive video generation
On CogenPVG (arXiv 2026), Chen is a middle author on a generator-critic reflective multi-agent pipeline grounded in the Elaboration Likelihood Model, spanning storyboard planning and persuasion evaluation. Evolve credits the paper with a novel task framing but finds the human-study evidence weak — κ=0.33 with no significance tests — and rates reproducibility readiness low.

Timeline

2026Last author on "Dual-Frontier: When Can an Agent Trust Its World Model?" (arXiv), whose non-identifiability theory and certified verify-then-promote gates are reviewed by Evolve.
2026Middle author on "CogenPVG", framed by Evolve as the first ELM-guided multi-agent framework for persuasive video generation (arXiv).
2026OpenAlex snapshot records 3 works dated 2026, 13 distinct co-authors, and an even 50/50 split between LLM-agent research and generative-AI multimedia — the only indexed year in this profile so far.

Papers in the Evolve catalog

Frequent co-authors

Huatai Zhu×1 Qiang Chen×1 Ziqian Kou×1 Wenhao Li×1 Fei Wang×1 Yichao Cao×1 Xiu Su×1 Yuntian Xiao×1 Shoulong Zhang×1 Wenfeng Song×1 Yan Wang×1 Yi Chen×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.

Bibliometrics are OpenAlex-resolved only (identity: openalex, no ORCID); "Yi Chen" is a very common name, so this profile is name-resolved from public bibliographic data.
No citation-impact data is available: h-index and i10-index are 0 and per-year counts cover only 2026 (3 works).
In-depth assessment covers only the two arXiv papers reviewed by Evolve; other indexed works are not summarized here.
The reproducibility score (mean 38) rests on just 2 reviewed papers, so it is a coarse estimate.