Continual learning researcher working on brain-inspired modularity and pretrained-model representation refinement.
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—
Insufficient data: no publication-count or venue metrics are available yet. This axis would measure total research output — number of papers, authorship positions, and venues over time. It will be scored once a fuller bibliometric record exists.
Impact—
Insufficient data: no citation counts, h-index, or influence signals were provided. This axis would measure how widely the work is cited and used by other researchers, including citation structure and downstream adoption.
Focus0/100
The radar focus score is 0, computed from a field distribution split exactly 50/50 between 'machine learning (continual learning and representation learning)' and 'machine learning (NeuroAI / continual learning)'. In substance the two labels describe one thematically tight program — both center on continual learning — but the formal metric reads the even split as zero concentration, so the score reflects labeling breadth rather than genuine topic drift.
Collaboration60/100
The collaboration score is 60, with 19 distinct coauthors recorded across the reviewed bibliometric data. Frequent collaborators include Kanglei Zhou, Qi Cheng, Hongwei Yan, Liyuan Wang, Shuang Cui, Hang Su, Jun Zhu, Yi Zhong, Weiyi Dong, Chunyan Lan, and Jun Zhou. Authorship within the catalog spans first author (MePo++) and middle positions, consistent with mid-sized team science rather than solitary or very large-consortium work.
Activity—
Insufficient data: only two 2026 arXiv preprints are in scope, with no multi-year publication history. This axis would measure research cadence — papers per year, recency, and sustained productivity over time. It remains unscored until a longer record exists.
Reproducibility78/100
The reproducibility score is 78, computed as the mean over 2 reviewed papers. Paper-level readiness differs: MePo++ is rated high, supported by broadly consistent reported gains, while the brain-inspired hierarchical modularity paper is rated medium, with strong empirical results but informal theory. This makes reproducibility the strongest currently scored axis, though it rests on only two preprints.
Research summary
Guanglong Sun is a machine learning researcher whose reviewed work centers on general continual learning — the problem of acquiring new tasks without catastrophic forgetting. The Evolve catalog holds two 2026 arXiv papers. The first-authored MePo++ unifies representation refinement and reconciliation for continual learning with pretrained models, combining meta-refinement with online alignment; Evolve reports broad, consistent benchmark gains while cautioning that representation-level effect sizes are very small, and rates reproducibility readiness high. The co-authored Brain-Inspired Hierarchical Modularity paper transfers organizational principles from the Drosophila mushroom body into a mixture-of-experts plus ensemble design, yielding large continual-learning gains in vision and robotics with strong empirics, informal theory, and replay-competitive performance; its reproducibility readiness is medium. Bibliometric raw data record 19 distinct coauthors and a field distribution split 50/50 between 'machine learning (continual learning and representation learning)' and 'machine learning (NeuroAI / continual learning)' — two labels for one thematically tight program. Radar scores place collaboration at 60 and reproducibility at 78, the mean over those 2 papers, while output, impact, and activity lack sufficient data. With both papers being recent preprints, this profile should be read as an early-stage snapshot of an active, tightly focused research line.
Collaboration pattern
Work in the Evolve catalog is collaborative: 19 distinct coauthors appear in the bibliometric record, and frequent collaborators include Kanglei Zhou, Qi Cheng, Hongwei Yan, Liyuan Wang, Shuang Cui, Hang Su, Jun Zhu, Yi Zhong, Weiyi Dong, Chunyan Lan, and Jun Zhou. The pattern is mid-sized team science — first-authoring MePo++ while contributing in a middle position to the brain-inspired modularity paper, both within the continual-learning and NeuroAI space.
Research keywords
Highlights
First-authored MePo++ unifies representation refinement and reconciliation for PTM-based continual learning
Co-authored brain-inspired hierarchical modularity for general continual learning
Timeline
- 2026Published first-authored MePo++ on arXiv, unifying representation refinement and reconciliation for general continual learning with pretrained models.
- 2026Co-authored the brain-inspired hierarchical modularity paper, applying Drosophila mushroom body-inspired mixture-of-experts and ensemble design to continual learning in vision and robotics.
- 2026Both papers entered the Evolve review catalog, with reproducibility readiness rated high (MePo++) and medium (brain-inspired modularity) and a mean reproducibility score of 78.
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 and radar scores derive from OpenAlex-style public bibliometric snapshots; no citation, h-index, publication-volume, or activity metrics were available, so output, impact, and activity remain unscored.
- Profile is name-resolved from public bibliographic data; no ORCID or institutional affiliation was provided, so identity disambiguation for this name is not guaranteed.
- Coverage is limited to 2 arXiv preprints (both 2026) reviewed by Evolve; peer-reviewed papers, other venues, and non-cataloged work are not reflected.
- The reproducibility score of 78 is a mean over only 2 papers, and Evolve's review flags very small representation-level effect sizes in MePo++.