Machine learning researcher advancing brain-inspired and pretrained-model approaches to general continual learning
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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: publication counts, venue types, and first-author shares are not available in the underlying data for this profile. This axis would measure total research production and the balance between leading and supporting roles across all recorded outputs.
Impact—
No impact score yet: citation counts, aggregate citation-style indicators, and downstream uptake are absent from the current record. This axis would measure how much later research builds on and references these two reviewed papers.
Focus0/100
Scores 0/100 on the composite focus axis, the lowest band. The underlying field distribution splits evenly: 50% machine learning (continual learning and representation learning) and 50% machine learning (NeuroAI / continual learning). In substance the two labeled subfields overlap heavily around continual learning, so the floor score reflects the even split across two field tags rather than a genuinely scattered agenda.
Collaboration60/100
Scores 60/100. The record shows 19 distinct coauthors across the two reviewed 2026 papers, with a recurring core including Guanglong Sun, Kanglei Zhou, Qi Cheng, Liyuan Wang, Shuang Cui, Hang Su, Jun Zhu, Yi Zhong, Weiyi Dong, Chunyan Lan, and Jun Zhou. Hongwei Yan led one paper as first author and sat in a middle position on the other, indicating both leadership and supporting roles within a moderately sized, concentrated network.
Activity—
No activity score yet: publication cadence and recency cannot be computed from the available record, which centers on 2026 arXiv releases. This axis would measure sustained publication output and engagement over recent years.
Reproducibility78/100
Scores 78/100, computed as the mean reproducibility-readiness rating over 2 reviewed papers. The brain-inspired modular architecture paper is rated medium readiness, consistent with heavier engineering and compute demands, while MePo++ is rated high, matching its broad benchmark reporting on pretrained models. Together they suggest a solid reproducibility habit for an early-stage record, though a two-paper evidence base remains thin.
Research summary
Hongwei Yan is a machine learning researcher whose documented record is early-stage, centered on general continual learning — learning new tasks without catastrophic forgetting — approached from two complementary directions. The first is brain-inspired architecture design: a 2026 first-author arXiv paper, Brain-Inspired Hierarchical Modularity for General Continual Learning, adapts the Drosophila mushroom body into a hierarchical mixture-of-experts-plus-ensemble design, reporting large continual-learning gains in vision and robotics; Evolve's review finds the empirics strong and replay-competitive while noting the theory remains informal. The second is representation-level methods for pretrained models: MePo++ (2026) unifies representation refinement and reconciliation through meta-refinement and online alignment, reporting broad consistent gains, though representation-level effect sizes are very small. The two papers split evenly across two closely related fields — machine learning (continual learning and representation learning) and machine learning (NeuroAI / continual learning) — and mean reproducibility readiness is 78/100, with MePo++ rated high and the brain-inspired paper medium. The collaboration footprint spans 19 distinct coauthors, including recurring collaborators such as Guanglong Sun, Kanglei Zhou, Liyuan Wang, Hang Su, and Jun Zhu; the collaboration dimension scores 60/100. This full but early-stage profile has no institution, country, or ORCID on record, and the output, impact, and activity axes lack sufficient data; identity is name-resolved from public bibliographic data.
Collaboration pattern
The record spans 19 distinct coauthors across two 2026 papers. Hongwei Yan led the brain-inspired paper as first author and contributed in a middle position to MePo++, showing both leadership and supporting roles. Frequent collaborators include Guanglong Sun, Kanglei Zhou, Qi Cheng, Liyuan Wang, Shuang Cui, Hang Su, Jun Zhu, Yi Zhong, Weiyi Dong, Chunyan Lan, and Jun Zhou, indicating a concentrated, recurring core team around continual-learning work.
Research keywords
Highlights
Drosophila-inspired hierarchical modularity for continual learning
MePo++: refining and reconciling pretrained representations
A reproducibility-forward early track record
Across the two reviewed 2026 papers, mean reproducibility readiness is 78/100 — MePo++ rated high and the brain-inspired paper medium — while the collaboration footprint spans 19 distinct coauthors, giving a collaboration score of 60/100 on an otherwise early-stage record.
Timeline
- 2026Released the first-author arXiv paper 'Brain-Inspired Hierarchical Modularity for General Continual Learning', applying Drosophila mushroom-body principles to general continual learning in vision and robotics.
- 2026Co-authored 'MePo++: Unifying Representation Refinement and Reconciliation for General Continual Learning' in a middle author position, targeting continual learning with pretrained models.
- 2026Both 2026 papers entered the Evolve review catalog, establishing a 19-coauthor collaboration footprint and a mean reproducibility readiness of 78 for this early-stage profile.
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.
- Identity is name-resolved from public bibliographic data; 'Hongwei Yan' is a common name and no ORCID, institution, or country is on record.
- Radar scores for output, impact, and activity are missing because publication-volume and citation metrics are not yet available in the underlying data.
- The evidence base is two 2026 arXiv preprints reviewed by Evolve; no peer-reviewed venues or pre-2026 publications are captured.
- The supplied frequent-coauthor list includes 'Hongwei Yan' itself, likely a name-disambiguation artifact in the source data.