NTU researcher in LLM-agent security, cost-aware agent efficiency, LLM reasoning, and federated graph learning.
4
papers here
4
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
27
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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 publication-output data Evolve needs is not available for this profile yet. This axis would measure aggregate productivity — total works, output trajectory across years, and venue-normalized volume — and will be scored once complete indexing coverage is obtained.
Impact—
Not yet scored: citation-based impact data is insufficient for this axis at present. It would measure influence — citations, citation trajectory over time, and uptake of the reviewed papers — and will be computed when adequate bibliometric coverage becomes available.
Focus0/100
Focus scores 0, reflecting an exceptionally dispersed portfolio rather than a thin record: the current field distribution splits evenly at 0.25 across efficient LLM agents, federated graph learning, LLM reasoning, and prompt-injection security, while topic assignments span five areas (privacy-preserving technologies 59, IoT/edge 42, blockchain 20, UAV 18, domain adaptation 17). The breadth is genuine, but no single field dominates the recent catalog.
Collaboration51/100
Collaboration scores 51, mid-range for this profile. The recent three-year record lists 58 works with 12 distinct coauthors, and the recurring circle is compact: Tiantong Wu is the only name entered twice in the frequency list, followed by Shutong Zheng, Lele Fu, Sheng Huang, Chuan Chen, Zhiren Gong, Yikun Hou, Zihao Zeng, Ming Xiao, and Chau Yuen. On the four reviewed 2026 papers he takes the last-author seat three times and a middle seat once.
Activity100/100
Activity scores 100, the strongest axis here. OpenAlex counts 58 works over the last three years and a rising curve: 18 works in 2020, 22 in both 2021 and 2022, 13 in 2023, a dip to 9 in 2024, then 16 in 2025 and 33 in 2026 — the largest single-year figure in the window. All four Evolve-reviewed papers date from 2026.
Reproducibility49/100
Reproducibility scores 49, averaged over four assessed papers whose Evolve readiness ratings run low, low, medium, high. The two Jev-based agent papers rest on self-built benchmarks and a single proprietary model, State of Thought shows weakened baselines and untested statistics, while the SPIRE federated-learning paper rates high. Releasing artifacts and baselines is the clearest lever for improvement.
Research summary
Wei Yang Bryan Lim is a researcher at Nanyang Technological University, Singapore. His OpenAlex-linked record shows long-standing volume in privacy-preserving technologies (59 topic-assigned works), IoT and edge/fog computing (42), blockchain applications and security (20), UAV applications and optimization (18), and domain adaptation and few-shot learning (17), with an h-index of 27 and 49 i10-indexed publications. Recent output is strong: 58 works in the last three years and 33 in 2026 alone, the highest single-year figure in the provided window. His current focus has pivoted toward LLM agents. Evolve reviewed four 2026 arXiv papers: prompt-injection attacks on Jev's typed probabilistic decisions, REFLEX's cost-aware selective control (72.7% fewer strong-model calls at 95% success), a State of Thought hidden-state controller for reasoning, and structural-entropy-driven diffusion for one-shot federated graph learning. Across these he holds the last-author position three times and a middle position once, none marked corresponding — a pattern consistent with a senior supervisory role. Evolve's reproducibility assessment averages 49 over the four papers, with verdicts citing self-built benchmarks, a single proprietary model, weakened baselines, and missing error bars as the main limitations.
Collaboration pattern
Lim works within a compact, stable circle: the recent three-year window lists 58 works with only 12 distinct coauthors, and the frequency list is topped by Tiantong Wu — the only name recorded twice — followed by Shutong Zheng, Lele Fu, Sheng Huang, Chuan Chen, Zhiren Gong, Yikun Hou, Zihao Zeng, Ming Xiao, and Chau Yuen. On the reviewed papers he occupies the last-author position three times and a middle position once, and none is marked corresponding.
Research keywords
Highlights
REFLEX cuts strong-model calls by 72.7% at 95% success
Prompt injection shifts Jev's action probabilities but rarely hijacks its choices
State of Thought: a 582-parameter controller for endogenous reasoning
SPIRE brings structural entropy and diffusion to one-shot federated graph learning
Timeline
- 2020OpenAlex records 18 works in 2020, the earliest year in the provided counts-by-year window.
- 2022Output holds at 22 works in both 2021 and 2022, the sustained plateau of the tracked record.
- 2024The tracked count dips to 9 works in 2024, the lowest point of the window.
- 2025Output recovers to 16 works in 2025, beginning the climb toward the record high.
- 2026A record 33 works appear in 2026, including all four Evolve-reviewed papers spanning prompt-injection security, cost-aware agent routing, reasoning control, and federated graph learning.
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.
- Bibliometrics (h-index 27, i10-index 49, counts by year, topics) come from an OpenAlex-linked snapshot; output and impact dimension scores are not yet computed.
- Coauthor coverage reflects the recent three-year window only (12 distinct coauthors); longer-term collaboration history is outside this snapshot.
- The reproducibility mean of 49 is based solely on four 2026 arXiv papers reviewed by Evolve; other works are unassessed.
- Identity is name-resolved from public bibliographic data (OpenAlex; ORCID 0000-0003-2150-5561); no biographical details beyond the provided record are included.