Early-career USTC researcher working on continual LLM personalization and long-term parametric memory for LLM agents.
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
2
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-09) 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: only a thin publication record exists for this name (7 works indexed across three years), too little for a stable output score. This axis would measure total works, venue quality, and the output trajectory once indexing matures.
Impact—
Not yet scored: citation data is essentially absent for this early-career record (i10-index 0), so impact cannot be estimated. This axis would measure citations, citing works, and downstream uptake as the two 2026 arXiv papers accumulate attention.
Focus0/100
The focus score is 0, reflecting strong topical dispersion in the OpenAlex record: Topic Modeling (3 works) sits alongside particle physics (2), astrophysics and cosmic phenomena (2), multimodal machine learning (2), and cosmology and gravitation theories (1). The reviewed catalog itself is concentrated — field distribution splits evenly between LLM agent memory (0.5) and LLM personalization (0.5) — so the physics topics likely reflect name-disambiguation noise rather than genuine breadth.
Collaboration51/100
The collaboration score is 51, based on 12 distinct coauthors in the last three years. The frequent-coauthor list shows a small recurring core — Liang Dong, Jian Xu, and Guanjun Jiang each appear twice — alongside Fugen Yao, Fanyu Zhao, Han Zhang, Yifei Zhao, Yinsheng Li, and Li Xiao. Authorship roles span both modes: first author on COPE and middle author on RPMem.
Activity61/100
The activity score is 61, driven by an accelerating curve in the OpenAlex counts-by-year data: 1 work in 2024, 2 in 2025, and 4 in 2026, totalling 7 works over the last three years. Both Evolve-reviewed papers appeared in 2026, including a first-authored arXiv preprint, indicating a clearly rising research tempo.
Reproducibility55/100
The reproducibility score is 55, computed as the mean across 2 reviewed papers, both rated medium readiness. Evolve's verdicts flag simulator-coupled evaluation and modest scores for COPE, and multiple-choice-only evaluation with benchmark-dependent gains for RPMem — concrete limitations that temper otherwise well-ablated results.
Research summary
Ruike Cao is an early-career researcher at the University of Science and Technology of China working across two complementary directions in applied machine learning: continual personalization of large language models under sparse user feedback, and long-term parametric memory for LLM agents. The OpenAlex record (h-index 2, i10-index 0) is consistent with the start of an independent publishing career: 7 works indexed in the last three years, with output accelerating year over year (1 in 2024, 2 in 2025, 4 in 2026). Evolve has reviewed two 2026 arXiv papers. COPE, with Cao as first author, personalizes LLMs continually through per-user soft embeddings and self-calibrated proxy rewards; Evolve judges the gains real but simulator-coupled, with scores staying modest. RPMem, with Cao as a middle author, proposes a backbone-reusable recurrent parametric memory consolidated across sessions via hypernetwork-generated LoRA; gains are strong and well-ablated on PERMA but modest elsewhere, with multiple-choice-only evaluation. Both papers rate medium on reproducibility readiness (mean 55 across 2 papers). The catalog field distribution splits evenly between LLM agent memory (0.5) and LLM personalization (0.5). OpenAlex topics also list particle physics and astrophysics, most plausibly name-disambiguation noise rather than a second research line. This is a full but early-stage profile; output and impact remain unscored pending citation data.
Collaboration pattern
Collaboration is moderate (score 51), with 12 distinct coauthors over the last three years. The pattern shows a small recurring core — Liang Dong, Jian Xu, and Guanjun Jiang each appear twice in the frequent-coauthor list — supplemented by Fugen Yao, Fanyu Zhao, Han Zhang, Yifei Zhao, Yinsheng Li, and Li Xiao. Authorship spans both led and team work: Cao is first author on COPE and a middle author on RPMem, consistent with a junior researcher contributing within a group while increasingly leading.
Research keywords
Highlights
First-authored COPE: continual LLM personalization under sparse feedback
RPMem: recurrent parametric memory for long-term LLM agents
Timeline
- 2024First works appear in the OpenAlex counts-by-year record (1 work).
- 2025Indexed output doubles to 2 works, continuing the early ramp-up.
- 2026Most active year: 4 indexed works, including the two Evolve-reviewed arXiv papers — first-authored COPE on continual LLM personalization and RPMem on recurrent parametric agent memory.
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 come from OpenAlex only; the h-index (2) and i10-index (0) reflect that single indexed record and may undercount real output.
- OpenAlex topic counts include particle physics, astrophysics, and cosmology despite an all-ML reviewed catalog; this profile is name-resolved from public bibliographic data (ORCID 0000-0002-4184-3560), so some indexed works may belong to other researchers with the same name.
- Output and impact axes are unscored because citation data is insufficient; the reproducibility estimate rests on only 2 reviewed papers.
- Both reviewed papers are 2026 arXiv preprints; the provided data contains no peer-reviewed venue information.