Early-career Harbin Institute of Technology researcher on long-term memory and personalization for LLM agents
3
papers here
3
reviewed by Evolve
3
first-author
0
corresponding
—
works · OpenAlex
3
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-14) 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—
The output axis measures publication volume and trajectory. It is not yet scored: OpenAlex coverage of this early-career record is incomplete, so the indexed counts (2022: 2, 2023: 1, 2024: 1, 2026: 8) are treated as insufficient.
Impact—
The impact axis measures citation-based influence. No score yet: an h-index of 3 and i10-index of 1 on a 2022-start, arXiv-heavy record leave citation coverage too immature for Evolve to score.
Focus0/100
The focus score is 0: the field distribution splits exactly into thirds across conversational AI and agent memory, LLM agents with long-term memory and personalization, and conversational AI, with no single dominant subfield. Topic data is similarly broad — topic modeling (7), NLP techniques (4), speech and dialogue systems (4), problem solving and planning (3), sentiment analysis (2) — though the three 2026 first-author papers converge on one theme: memory for conversational LLM agents.
Collaboration62/100
The collaboration score is 62, derived from 21 distinct coauthors in the indexed record, 12 of whom recur across the reviewed work, including Jiangyue Yan, Zike Yuan, Ruifeng Xu, Xiaoyu Shen, and Junnan Dong. All three catalog papers are first-authored and none is corresponding, consistent with a junior researcher contributing the lead experimental work within a rotating multi-author group.
Activity67/100
The activity score is 67. The indexed record shows 9 works in the last three years, with countsByYear of 2 (2022), 1 (2023), 1 (2024), and 8 (2026) — a sharp 2026 acceleration in which all three Evolve-reviewed first-author arXiv papers on agent memory appeared.
Reproducibility20/100
The reproducibility score is 20, computed from the three reviewed papers (reproMean 20, reproPapers 3). All three rate low: the latent neuro-symbolic memory paper lacks significance tests, a stated judge, and code; the interactive memory RL paper shows LLM-judge circularity and thin reproducibility; ThinkFlow contains duplicated ablation rows and closed-source overclaims.
Research summary
Cai Ke is an early-career natural language processing researcher at Harbin Institute of Technology whose profile is name-resolved from OpenAlex bibliographic data (ORCID 0009-0008-2207-7304). The indexed record begins in 2022 and shows 9 works in the last three years, with a sharp step-up in 2026 (8 works recorded that year). The research program centers on memory for LLM-based conversational agents: all three papers Evolve has reviewed are 2026 first-author arXiv preprints — latent neuro-symbolic memory disentangling, interactive memory learning via two-agent reinforcement learning, and a self-evolving probabilistic latent memory system for lifelong agents. Bibliometric topics back this up: topic modeling (7), NLP techniques (4), speech and dialogue systems (4), AI-based problem solving and planning (3), and sentiment analysis (2). Field distribution splits evenly across three NLP subfields covering agents, memory, and conversational AI. The scholarly record is young — h-index 3, i10-index 1 — and Evolve's reviews flag recurring evaluation and reproducibility gaps (missing significance tests, LLM-judge circularity, duplicated ablation rows, no code release), making reproducibility the clearest near-term lever for strengthening the work. Output and impact radar axes remain unscored pending better coverage.
Collaboration pattern
The indexed record spans 21 distinct coauthors, with a recurring group of 12 including Jiangyue Yan, Zike Yuan, Han Zhang, Xin Liu, Ruifeng Xu, Xiaoyu Shen, Keyu Chen, and Junnan Dong. All three Evolve-reviewed papers are first-authored, non-corresponding preprints with overlapping keyword sets — a pattern consistent with a junior researcher leading experimental memory-agent work inside a collaborative NLP group with rotating coauthor combinations.
Research keywords
Highlights
Query-conditioned latent graphs for disentangled long-term memory
Reinforcement learning for memory management in long-term conversations
Self-evolving probabilistic latent memory for lifelong agents
Timeline
- 2022Earliest indexed works appear, with 2 works recorded in the OpenAlex counts-by-year data.
- 2023One work recorded in the indexed year.
- 2024One work recorded; the indexed pace stays low before the later ramp-up.
- 2026Productivity accelerates sharply: 8 works recorded, including three first-author arXiv preprints on LLM agent memory that enter the Evolve catalog.
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 come from OpenAlex only (identity: openalex); coverage from Scopus, Web of Science, or Semantic Scholar is not reflected.
- The output and impact radar axes are unscored due to insufficient coverage; the h-index of 3 and i10-index of 1 reflect a record starting in 2022 with citation lag on 2026 arXiv preprints.
- Only the three 2026 arXiv preprints listed have full Evolve reviews; verdicts may change after peer review.
- "Cai Ke" is a common Chinese name and the Chinese-character spelling is unconfirmed; this profile is name-resolved from public bibliographic data (ORCID 0009-0008-2207-7304) and may not cover every same-name publication.