LLM-agent researcher listed at UIC: factuality verification, self-evolving time-series agents, agent memory
3
papers here
2
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—
Output is not scored yet. This axis would measure publication volume and productivity trajectory (works per year, venue spread) from a consolidated author record; the current OpenAlex profile appears fragmented, so no reliable output figure can be reported.
Impact—
Impact is not scored yet. This axis would measure citation-based influence (citations, h-index, i10-index). The record lists h-index 1 and i10-index 0 against 139 logged works — a contradiction that signals incomplete author disambiguation rather than a true impact level.
Focus0/100
Focus scores 0, reflecting a deliberately split portfolio: OpenAlex field tagging divides evenly at 0.5/0.5 between NLP factuality evaluation/LLM agents and machine learning for LLM time-series agents. Tracked topics spread across at least five areas — topic modeling (59 mentions), advanced graph neural networks (32), domain adaptation and few-shot learning (24), NLP techniques (21), and multimodal applications (21) — with the leading topic at roughly 38% of 157 topic mentions.
Collaboration79/100
Collaboration scores 79, based on 51 distinct coauthors in the record. In the Evolve catalog this author holds the last position on two of three reviewed 2026 papers (TimeEvo, EAVer) and a middle position on Hindsight Memory-PRM, a pattern consistent with a senior or supervisory slot. Twelve recurring coauthors anchor the network, including Kening Zheng, Aoying Zheng, Wei-Chieh Huang, Yazhi Guo, and Zhigang Chang.
Activity100/100
Activity scores at the maximum of 100. The record logs 111 works over the trailing three years, and the year-by-year counts jump from 11 in 2024 and 12 in 2025 to 88 dated 2026 — the largest single year in a series that also includes 11 in 2021, 7 in 2022, and 4 in 2023. The current publication tempo is exceptionally dense relative to the record's own history.
Reproducibility55/100
Reproducibility scores 55, the mean across the 2 papers scored so far (reproPapers: 2). Both reviewed papers carry a medium repro-readiness rating. Evolve's verdict on EAVer flags that no code is released and no variance is reported, while TimeEvo is called rigorous in protocol but is run without self-evolving baselines — concrete gaps that cap the score at mid-range.
Research summary
This profile is name-resolved from a public OpenAlex author record listed under Philip S. Yu with a University of Illinois Chicago affiliation; because this name is shared by a very senior researcher, all figures here describe only the underlying record. The reviewed work converges on LLM-based agents. In two 2026 arXiv papers this author holds the last position: TimeEvo, which evolves a time series agent through failure-driven synthesis of a tool library under a paired admission gate, and EAVer, which trains an 8B agentic verifier for long-form factuality as an end-to-end policy using privileged-teacher trajectories and direct preference optimization. A third 2026 paper, Hindsight Memory-PRM, supervises agent memory management with auditable hindsight credit, with this author in a middle position. OpenAlex field tagging splits evenly, 0.5/0.5, between NLP factuality evaluation/LLM agents and machine learning for LLM time-series agents, while the largest tracked topics are topic modeling (59 mentions), advanced graph neural networks (32), and domain adaptation and few-shot learning (24). Output volume is high in the record: 111 works in the trailing three years, including 88 dated 2026, with 51 distinct coauthors. Reproducibility sits mid-range at a mean of 55 across 2 scored papers, both rated medium readiness.
Collaboration pattern
The record lists 51 distinct coauthors, anchored by twelve recurring collaborators: Kening Zheng, Aoying Zheng, Zhigang Chang, Yazhi Guo, Miaotian Guo, Qingwei Zong, Xianhai Xie, Weiqiang Jin, Chengze Li, Hanrong Zhang, Jie Yang, and Wei-Chieh Huang. Within the Evolve catalog this author takes the last position on two of the three 2026 papers (TimeEvo, EAVer) and a middle position on the third, a placement pattern consistent with a senior or supervisory authorship role.
Research keywords
Highlights
Failure-driven self-evolution for a time series agent
Long-form factuality verification as an agentic policy
Auditable hindsight credit for agent memory management
2026 arXiv paper with this author in a middle position, supervising memory management in agents through hindsight credit assignment; it has not been reviewed by Evolve yet, so no verdict, keywords, or repro-readiness score are available.
Timeline
- 2026TimeEvo: failure-driven self-evolution of a time series agent (arXiv; last author)
- 2026EAVer: long-form factuality verification as an end-to-end agentic policy (arXiv; last author)
- 2026Hindsight Memory-PRM: supervising memory management with auditable hindsight credit (middle author)
- 2026OpenAlex logs 88 works dated 2026 in this record — the largest single year among the provided counts
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.
- Bibliometric figures are computed from a single OpenAlex author record (identity: openalex); no ORCID linkage was provided for verification.
- Philip S. Yu is a widely shared name: this profile is name-resolved from public bibliographic data only, and the record's h-index (1) and i10-index (0) conflict with its 139 logged works, indicating likely incomplete author disambiguation.
- The source lists country code CN alongside a University of Illinois Chicago affiliation; this inconsistency is unresolved in the provided data.
- Only 3 papers from this author are in the Evolve review catalog, 2 with verdicts; Hindsight Memory-PRM lacks a verdict, keywords, and slug, and output/impact dimensions are unscored.