Fudan University researcher working on memory for LLM agents — calibration benchmarks and recurrent parametric memory.
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
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—
No publication-volume data is available in this snapshot, so output is not scored. This axis would measure total scholarly production — number of papers, venue types, and authorship positions — across the researcher's full record.
Impact—
Citation and uptake metrics are not available, so impact is not scored. This axis would measure how widely the two catalogued papers are cited, reused, or built upon by other LLM-agent memory work.
Focus0/100
The radar scores focus at 0, derived from the field distribution: output splits evenly at 0.5/0.5 between "machine learning / NLP (LLM agent memory)" and "machine learning (LLM agents / NLP)". Both labels describe essentially the same topic — memory in LLM agents — so the low score likely reflects label granularity rather than genuine topic scattering across the two catalogued 2026 papers.
Collaboration50/100
Collaboration scores 50, based on 11 distinct co-authors across the catalogued record. Recurring names include Fanyu Zhao, Liang Dong, Jian Xu, Guanjun Jiang, Han Zhang, Ruike Cao, Fugen Yao, Yinsheng Li, and Li Xiao. On both 2026 papers Yifei Zhao holds a middle authorship position without corresponding-author role, suggesting contribution within a team rather than solo leadership so far.
Activity—
Only two papers, both dated 2026, are present in this snapshot, so research activity over time is not scored. This axis would measure publishing cadence and year-over-year volume, which requires a longer publication history than the catalogued record provides.
Reproducibility55/100
Reproducibility scores 55, the mean readiness across 2 assessed papers, both rated medium. Evolve's reviews note rigorous, well-ablated experiments in both: MemCalib's benchmark-plus-RL results are careful, and RPMem's gains are well ablated, though its evaluation is multiple-choice-only and MemCalib relies on an LLM-judge loop — caveats that cap confidence in replication.
Research summary
Yifei Zhao is a researcher at Fudan University whose catalogued work centers on memory for large language model (LLM) agents. The Evolve snapshot contains two 2026 arXiv papers, both with a middle authorship position and neither as corresponding author. The first, MemCalib, contributes a 15k-example benchmark for memory calibration in LLM agents paired with counterfactual-credit reinforcement learning; Evolve's review calls it rigorous and well-ablated while flagging that LLM-judge circularity weakens some conclusions. The second, RPMem, proposes a backbone-reusable recurrent parametric memory consolidated across sessions using hypernetwork and LoRA components; it reports strong, well-ablated gains on PERMA but only modest improvements elsewhere, and its evaluation is multiple-choice-only. Bibliometrically, the record shows 11 distinct co-authors, and the field distribution splits evenly (0.5/0.5) between "machine learning / NLP (LLM agent memory)" and "machine learning (LLM agents / NLP)" — two labels describing the same LLM-agent-memory topic. Radar scores are collaboration 50, reproducibility 55 (mean readiness across 2 assessed papers, both medium), and focus 0; output, impact, and activity are unscored for lack of data. Frequent co-authors include Fanyu Zhao, Liang Dong, Jian Xu, Guanjun Jiang, Han Zhang, Ruike Cao, Fugen Yao, Yinsheng Li, and Li Xiao. Because "Yifei Zhao" is a common name, this profile is name-resolved from public bibliographic data.
Collaboration pattern
Yifei Zhao's catalogued work is team-based: 11 distinct co-authors appear across the record, with repeat collaborators including Fanyu Zhao, Liang Dong, Jian Xu, Guanjun Jiang, Han Zhang, Ruike Cao, Fugen Yao, Yinsheng Li, and Li Xiao. Both 2026 papers place Yifei Zhao in a middle authorship position without corresponding-author status, indicating a contributing role within an LLM-agent memory group at Fudan University rather than paper leadership so far.
Research keywords
Highlights
MemCalib: a 15k-example benchmark plus counterfactual-credit RL for memory use in LLM agents
RPMem: recurrent parametric memory consolidated across sessions
Timeline
- 2026MemCalib appears on arXiv — a 15k-example memory-calibration benchmark for LLM agents combined with counterfactual-credit reinforcement learning.
- 2026RPMem appears on arXiv — recurrent parametric memory for LLM agents, consolidated across sessions with hypernetwork and LoRA components.
- 2026Both catalogued papers land in the same year with middle authorship and no corresponding-author role, involving 11 distinct co-authors.
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 signals come from the platform's OpenAlex-scope snapshot only; citation counts, h-index, and earlier publication history are not included.
- Name resolution: "Yifei Zhao" is a common name, so this profile is resolved from public bibliographic data and may merge or split identities.
- Only two 2026 arXiv preprints were reviewed; peer-reviewed venue acceptance is not reflected here.
- Output, impact, and activity radar axes are unscored due to missing data; focus is scored 0 from a 0.5/0.5 field-label split that may reflect label granularity rather than real topic dispersion.