Machine learning researcher in continual chunking and predictive coding networks; both 2026 arXiv papers first-authored.
2
papers here
1
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) 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 metrics are available yet. This axis would measure publication volume and productivity — total paper counts, output rate, and venue spread over time. With only a two-paper reviewed sample, no score is assigned.
Impact—
Impact data is not available yet. This axis would measure citation counts, citation velocity, and downstream influence on follow-up work. The current snapshot contains no citation or h-index figures, so the axis is left unscored.
Focus100/100
Radar focus scores 100/100: the bibliometric field distribution collapses to a single labeled field — machine learning (self-organizing systems / continual learning) — and both cataloged 2026 papers sit in adjacent ML subfields sharing one methodological motif, chunking. Sheaf SyncMap targets stable continual chunking; the predictive-coding study tunes the order of hidden-state optimization. Within this small snapshot the agenda reads as deliberately narrow rather than scattered.
Collaboration22/100
Collaboration scores 22/100, reflecting a very small footprint: raw data lists 2 distinct coauthors across the snapshot, with Danilo Vasconcellos Vargas the repeated collaborator on both 2026 papers as listed among frequent coauthors. Both papers are first-authored and non-corresponding. This is a tight, sustained two-person pattern rather than a broad network; counts come only from the reviewed catalog, so wider co-authorship may exist.
Activity—
Activity metrics are unavailable. This axis would measure publishing cadence — recency, frequency, and consistency of output across years. Only two 2026 papers are recorded, too few to characterize a temporal pattern, so it stays unscored.
Reproducibility55/100
Reproducibility scores 55/100 (reproMean 55), computed from 1 reproducibility-scored paper in the sample. Both cataloged papers carry a medium reproducibility readiness rating. The Sheaf SyncMap paper reports strong synthetic gains (17/18 graphs), yet its evaluation is oracle-tuned with SyncMap-only baselines; the predictive-coding study reports +5.5–9.8pp on CIFAR-10 but is MLP-only. Medium readiness plus these scope limits suggests replication is feasible but should check setup dependencies.
Research summary
Xueyuan Li is a machine learning researcher whose reviewed record currently consists of two first-authored 2026 arXiv papers, both written with Danilo Vasconcellos Vargas within a very small collaboration footprint of 2 distinct coauthors. The first paper studies hidden-state optimization order in predictive coding networks: its boundary-first chunked inference lifts CIFAR-10 accuracy over standard predictive coding by 5.5–9.8pp, though it remains an MLP-only empirical study of modest scope. The second applies sheaf theory — Sheaf SyncMap with sheaf-Laplacian regularization — to unsupervised continual chunking, stabilizing performance with strong synthetic gains on 17/18 graphs, while oracle-tuned evaluation and SyncMap-only baselines temper the claims. Both papers carry a medium reproducibility readiness rating, and the reproducibility dimension scores 55 from a single scored paper. Focus scores 100, reflecting a single labeled field and one unifying motif — chunking, whether as a continual-learning task or as an inference schedule. Output, impact, and activity metrics are not yet available, so productivity and influence cannot be judged from this snapshot.
Collaboration pattern
The snapshot shows an unusually tight collaboration pattern: 2 distinct coauthors total, with Danilo Vasconcellos Vargas appearing repeatedly and credited on both 2026 papers. The researcher takes first authorship on both, without corresponding-author designation. The pattern suggests a mentoring or lab-based two-person collaboration rather than a distributed network. Because co-author counts come only from the reviewed catalog and bibliometric snapshot, broader collaborations outside this sample cannot be ruled out.
Research keywords
Highlights
Sheaf SyncMap stabilizes unsupervised continual chunking
Boundary-first inference lifts predictive coding accuracy
Chunking as a unifying research thread
Across both cataloged papers, chunking is the shared motif — as a continual-learning task stabilized by sheaf-based self-organization in one, and as the unit of an inference schedule in predictive coding networks in the other. The single-field focus distribution (focus score 100) reinforces this reading of a coherent, deliberately narrow agenda.
Timeline
- 2026First-authored arXiv paper on hidden-state optimization order in predictive coding networks; boundary-first chunked inference improves CIFAR-10 accuracy by 5.5–9.8pp over standard predictive coding.
- 2026First-authored arXiv paper on Sheaf SyncMap for unsupervised continual chunking; strong synthetic gains on 17/18 graphs, with medium reproducibility readiness.
- 2026Both papers are reviewed into the Evolve catalog; the snapshot records 2 distinct coauthors, with Danilo Vasconcellos Vargas as the recurring collaborator.
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.
- Radar scores and collaboration counts derive from a public bibliometric snapshot covering only the reviewed catalog; they may not reflect this researcher's full publication record.
- Both in-scope papers are 2026 arXiv preprints; no peer-reviewed venue record, citation counts, or h-index are available, so the output and impact axes are left unscored.
- The reproducibility score (55) is averaged over a single scored paper (reproPapers = 1) and should be read as preliminary.
- No institution, country, or ORCID is attached; identity is name-resolved from public bibliographic data, and 'Xueyuan Li' is not a rare name.