Tsinghua researcher advancing continual learning from brain-inspired modularity to egocentric video perception
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 output metric is available in this snapshot: publication counts, venue mix, and yearly volume are not yet computable from current sources. This axis would measure the volume and consistency of published output over time.
Impact—
Impact data is missing here: citation counts, h-index, and venue influence are not available in the current snapshot, so no score is given. This axis would measure how widely the researcher's papers are cited and taken up by the community.
Focus0/100
The focus score is 0, the lowest possible, because the cataloged field distribution splits evenly: 50% computer vision / machine learning work on continual learning for embodied multi-view video perception, and 50% machine learning work on NeuroAI and continual learning. In practice the split is more about application surface than topic — both halves center on continual learning — but the even 0.5/0.5 split reads as low concentration for now.
Collaboration51/100
Collaboration scores 51, a moderate level, with 12 distinct coauthors recorded. The frequent-coauthor list includes Hongwei Yan, Kanglei Zhou, Yi Zhong, Yuchen Liu, Qingyu Shi, and Qi Cheng, among others. Both cataloged papers place him in the last (senior) author position rather than as corresponding author, suggesting a senior supervisory role within a moderately sized, stable collaboration network.
Activity—
No activity score is computed: recent publication cadence, preprint postings, and year-over-year engagement cannot be derived from the current data. This axis would measure how actively the researcher publishes and engages with the field right now.
Reproducibility55/100
Reproducibility averages 55 across the 2 papers Evolve has assessed, both rated medium readiness. For the brain-inspired modularity work, reviewers credit strong empirics but informal theory; for CE4L, they judge results largely convincing while flagging seed variance. The moderate mean reflects empirical strength tempered by unverified robustness, and the score rests on a small two-paper base.
Research summary
Liyuan Wang, affiliated with Tsinghua University, works on continual learning along two complementary lines that his bibliographic field distribution splits evenly (50% / 50%): machine learning in the NeuroAI tradition, and computer vision for embodied multi-view video perception. The Evolve catalog contains two 2026 arXiv preprints, in both of which he sits in the last (senior) author position, though neither lists him as corresponding author. The first, 'Brain-Inspired Hierarchical Modularity for General Continual Learning', borrows organizational principles from the Drosophila mushroom body, combining mixture-of-experts and ensemble modularity; Evolve's review credits it with large continual-learning gains in vision and robotics, strong empirics, and informal theory, with results that remain competitive with replay baselines. The second, 'CE4L: Continual Ego, Exo, and Ego-Exo Learning', contributes a multi-view continual-learning benchmark plus training-free subspace-routed adapters; reviewers judge it largely convincing but flag router overclaim and seed variance. The collaboration radar records 12 distinct coauthors and a moderate collaboration score of 51; reproducibility averages 55 across the 2 papers assessed, both rated medium readiness. Output, impact, and activity metrics are not yet available in this snapshot, so those radar axes are left unscored.
Collaboration pattern
Liyuan Wang's collaboration footprint is moderate (score 51) with 12 distinct coauthors. Recurring partners include Hongwei Yan, Kanglei Zhou, Yi Zhong, Yuchen Liu, Qingyu Shi, Qi Cheng, and Weiyi Dong. On both 2026 arXiv papers he occupies the last-author slot — typically the senior supervisory seat — while neither paper designates him corresponding author, so formal correspondence likely routes elsewhere. The network skews toward a stable, mid-sized team rather than a wide web of external collaborations.
Research keywords
Highlights
Drosophila-inspired modular architecture delivers large continual-learning gains
CE4L: a continual ego-exo benchmark with training-free adapters
A dual-track continual-learning agenda spanning NeuroAI and embodied vision
Across both 2026 catalog entries his field distribution splits evenly (50%/50%) between NeuroAI-flavored machine learning and continual learning for embodied multi-view video perception, with him in the last-author seat on each — one thread importing biological circuit principles, the other stress-testing continual learning on realistic egocentric video.
Timeline
- 2026Senior (last) author of 'Brain-Inspired Hierarchical Modularity for General Continual Learning' on arXiv, applying Drosophila mushroom-body-inspired mixture-of-experts and ensemble modularity to continual learning in vision and robotics.
- 2026Senior (last) author of 'CE4L: Continual Ego, Exo, and Ego-Exo Learning' on arXiv, releasing a continual multi-view benchmark and training-free subspace-routed adapters.
- 2026Both cataloged papers this year place him in the last-author position without corresponding-author designation, and his field distribution divides evenly (50%/50%) between NeuroAI-oriented and embodied-vision continual learning.
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 indicators come from public bibliographic data (OpenAlex-style) only; no citation counts, h-index, or output/activity metrics were available in this snapshot.
- 'Liyuan Wang' is a common name; this profile is name-resolved from public bibliographic data and anchored only to the listed Tsinghua University affiliation, with no ORCID linked.
- Only two 2026 arXiv preprints from the Evolve catalog back this profile; the reproducibility score (55) rests on just 2 assessed papers, a small base.
- Preprints may not yet be peer-reviewed or appear in their final venues; affiliation country is not recorded in the source data.