Alibaba researcher on memory and personalization for LLM agents: benchmarks, parametric memory, RL credit assignment.
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
2
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-13) 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—
No output score yet: the snapshot lacks a complete, citation-normalized publication count for this author. Only minimal signals exist (h-index 2, i10-index 2, and 4 works in the last three years), so overall research output cannot be fairly rated on this axis.
Impact—
No impact score yet: citation counts are unavailable in this snapshot beyond an h-index of 2 and i10-index of 2. This axis would measure citation uptake, influence on follow-up agent-memory research, and adoption of these methods — the data is insufficient to estimate.
Focus0/100
Scores 0. The field distribution splits evenly across three adjacent areas — machine learning/NLP (LLM agent memory), machine learning (LLM personalization), and machine learning (LLM agents/NLP) at ~0.33 each — while the topic list also spans topic modeling (3), transportation planning and optimization (3), transportation and mobility innovations (3), multimodal ML (2), and reinforcement learning in robotics (1). The 2026 papers are tightly themed; the dispersion partly reflects earlier work or name-resolution noise.
Collaboration62/100
Scores 62. OpenAlex records 21 distinct coauthors in the recent window — broad for an h-index-2 profile. The frequent-coauthor list is dominated by an Alibaba-affiliated cluster: Jian Xu, Guanjun Jiang, and Liang Dong each appear repeatedly, alongside Fanyu Zhao, Ruike Cao, Han Zhang, Yifei Zhao, Yinsheng Li, and Li Xiao. On all three catalog papers Yao is a middle author and not corresponding, indicating team-based rather than lead collaboration.
Activity47/100
Scores 47. Recent activity is modest but accelerating: 4 works in the last three years, all clustered in 2026 (countsByYear: 2026 = 4), versus 2 indexed works in 2020 and 1 in 2021, with 2022–2025 unindexed. The 2026 cluster contains all three Evolve-reviewed arXiv papers on agent memory and personalization, suggesting a concentrated, currently active phase within the Alibaba group.
Reproducibility55/100
Scores 55, the mean across 3 repro-assessed papers, each rated medium readiness by Evolve. Strengths noted: MemCalib is described as rigorous and well ablated, and RPMem shows strong, well-ablated gains on PERMA. Deductions: COPE's gains are simulator-coupled with modest scores, MemCalib's LLM-judge loop introduces circularity, and RPMem's evaluation is multiple-choice-only — yielding an above-average but not yet high reproducibility profile.
Research summary
Fugen Yao is a researcher affiliated with Alibaba Group whose recent work centers on memory and personalization for large language model (LLM) agents. Three 2026 arXiv papers reviewed by Evolve outline a coherent agenda: COPE introduces continual personalization of LLMs under sparse user feedback using per-user soft embeddings and self-evaluation; MemCalib contributes a 15,000-example benchmark for memory use in LLM agents paired with counterfactual-credit reinforcement learning; and RPMem proposes recurrent parametric memory consolidated across sessions via hypernetwork-generated LoRA adapters. On all three, Yao is a middle author rather than corresponding author, consistent with a specialist-contributor role inside a larger industrial research group. OpenAlex bibliometrics remain thin: an h-index of 2, an i10-index of 2, four works in the last three years, and works indexed only in 2020 (2), 2021 (1), and 2026 (4). Collaboration is broad relative to output — 21 distinct coauthors, with Jian Xu, Guanjun Jiang, and Liang Dong recurring most often. Evolve rates reproducibility readiness as medium for all three papers (mean reproducibility score 55), crediting ablation rigor while flagging simulator-coupled gains, LLM-judge circularity, and multiple-choice-only evaluation. This profile is name-resolved from public bibliographic data; no ORCID is provided.
Collaboration pattern
Yao collaborates within a sizable industrial cluster: 21 distinct coauthors are recorded, with Jian Xu, Guanjun Jiang, and Liang Dong recurring most frequently, alongside Fanyu Zhao, Ruike Cao, Han Zhang, Yifei Zhao, Yinsheng Li, and Li Xiao. On all three reviewed 2026 papers Yao is a middle author and not corresponding, fitting a specialist-contributor role in Alibaba's LLM agent-memory work rather than sole-lead projects.
Research keywords
Highlights
COPE: continual LLM personalization from sparse feedback
MemCalib: a 15k-example memory-calibration benchmark for LLM agents
RPMem: recurrent parametric memory consolidated across sessions
Timeline
- 20202 works indexed in OpenAlex for this author, the earliest year in the provided countsByYear record.
- 20211 work indexed — output thins to a single indexed publication in this year.
- 20264 works indexed, all in this single year, including the three Evolve-reviewed arXiv papers COPE, MemCalib, and RPMem on LLM agent memory and personalization.
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 OpenAlex only; no ORCID is provided, so identity is name-resolved from public bibliographic data.
- Output and impact scores are unavailable (dimMissing); the only citation signals present are h-index 2 and i10-index 2.
- countsByYear lists only 2020, 2021, and 2026 — the 2022–2025 gap may reflect incomplete indexing rather than actual inactivity.
- Topic labels such as transportation planning and topic modeling diverge from the 2026 LLM agent papers and may reflect name-resolution noise in the OpenAlex record.