Varuna De Silva
—
ML researcher working across vital-sign latent modeling and communication-efficient multi-agent reinforcement learning
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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—
Publication output cannot be scored yet: the snapshot contains only the two Evolve-reviewed 2026 arXiv papers, with no full publication list, venue history, or paper counts. This axis would measure total output volume, authorship roles, and productivity over the researcher's career.
Impact—
Citation-based impact is not yet available: no citation counts, h-index, or citing works appear in the snapshot, and both known papers are 2026 arXiv preprints. This axis would measure citations, scholarly influence, and uptake of the work by other researchers.
Focus0/100
The focus score is 0/100, reflecting an even 50/50 split of the reviewed output between two distant fields: machine learning for healthcare / physiological signal processing (the CVAE vital-sign latent modeling paper) and multi-agent reinforcement learning (the MARC communication paper). With only two reviewed papers divided exactly evenly, no specialization is detectable yet; the score could only rise once a deeper, single-field record accumulates.
Collaboration28/100
Collaboration scores 28/100, based on only 3 distinct co-authors across the reviewed record. The recurring names — Rafael Pina, Mindula Illeperuma, and Corentin Artaud — suggest a small, stable core team rather than an extended network, and on both papers De Silva sits in a middle authorship position without corresponding-author roles. Wider co-authorship would be needed to lift this score.
Activity—
Publishing activity over time cannot be scored: the snapshot places both known papers in 2026, with no earlier or later records, no submission cadence, and no venue timeline. This axis would measure annual output rhythm and sustained research engagement.
Reproducibility60/100
Reproducibility averages 60/100 across the 2 reviewed papers, and the split is instructive. The multi-agent communication paper (MARC) is rated high for reproducibility readiness, even though Evolve notes it lacks random-seed and confidence-interval reporting; the vital-sign latent modeling paper is rated low, compounded by a mean-distance evaluation metric the review calls gameable. The average therefore rests on just two data points.
Research summary
Varuna De Silva is an early-stage profile built from a thin bibliographic snapshot: two 2026 arXiv papers reviewed by Evolve, split evenly 50/50 between two fields — machine learning for healthcare / physiological signal processing, and multi-agent reinforcement learning. In the healthcare line, a conditional variational autoencoder with a latent classifier reshapes vital signs from healthy individuals toward clinical, ICU-style conditions; Evolve finds the ablations interesting but flags the mean-distance evaluation metric as gameable, and rates reproducibility readiness low. In the multi-agent line, MARC targets robust and efficient communication for multi-agent learning, adding an observation-prediction regularizer and DCT compression to attention-based communication; reported gains are significant, though Evolve criticizes the weak theoretical grounding and the absence of random-seed and confidence-interval reporting, while rating reproducibility readiness high. On both papers De Silva appears in a middle authorship position without corresponding-author credit, working with a compact set of 3 distinct co-authors — Rafael Pina, Mindula Illeperuma, and Corentin Artaud — which yields a collaboration score of 28/100. Reproducibility averages 60/100 across the 2 reviewed papers, and the even two-field split produces a focus score of 0/100. Citation-based impact, output volume, and multi-year activity metrics are not yet available for this researcher.
Collaboration pattern
Co-authorship in this snapshot is compact: only 3 distinct co-authors appear across the two reviewed papers, and the same recurring names — Rafael Pina, Mindula Illeperuma, and Corentin Artaud — likely form the working core of both projects. De Silva occupies a middle authorship position on each paper without corresponding-author credit. The pattern suggests a small, stable team collaborating across both of the researcher's active areas rather than an extended or international network; the collaboration score is correspondingly low at 28/100.
Research keywords
Highlights
Reshaping healthy vital signs toward clinical conditions with latent modeling
MARC: robust, compressed communication for multi-agent learning
Timeline
- 2026Publishes 'Representing Clinical Conditions on Vital Signs from Healthy Individuals using Latent Modeling' on arXiv, entering machine learning for healthcare and physiological signal processing.
- 2026Publishes 'Robust and Efficient Communication for Multi-Agent Learning' (MARC) on arXiv in multi-agent reinforcement learning, with reproducibility readiness rated high.
- 2026The two reviewed papers split evenly (50/50) across the two fields, with 3 distinct co-authors — Rafael Pina, Mindula Illeperuma, Corentin Artaud — and a mean reproducibility of 60/100.
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.
- Institutional affiliation, country, and ORCID are absent from the input data; this profile is name-resolved from public bibliographic records and should be verified against a unique author identifier.
- Bibliometrics come from a thin snapshot covering only the two Evolve-reviewed 2026 arXiv preprints; no peer-reviewed venue record is included.
- Output, impact (citations, h-index), and activity scores are missing because no publication counts, citation records, or multi-year timelines are available.
- Field, collaboration, and reproducibility scores rest on only 2 papers (reproMean 60 over 2 papers; 3 distinct co-authors) and may shift substantially as the record grows.