Varshini Elangovan
Carnegie Mellon University · United States
Early-career CMU researcher working across real-time AI safety interventions and efficient vision-language models
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
1
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—
Not yet scored: the indexed record is too thin to compute a productivity estimate — three works in the past three years (one in 2024, two in 2026). This axis would measure publication volume, consistency, and output growth over time.
Impact—
Not yet scored: citation signals are too early-stage to support a stable estimate (h-index 1, i10-index 1 on a very young record). This axis would measure how widely the work is cited and its influence across fields.
Focus0/100
The focus score is 0, reflecting an evenly split record: the bibliometric field distribution is 50% human-computer interaction (AI safety) and 50% computer vision (efficient vision-language models), and five indexed topics (e.g., colorectal cancer screening, 3D shape modeling) each carry a single work. At this early stage the profile reads as deliberately two-sided rather than concentrated; depth in either line is not yet established.
Collaboration54/100
Collaboration scores 54, based on 14 distinct coauthors in the indexed record and 12 named frequent coauthors (including Sauvik Das, Virginia Smith, and Rakesh Ranjan). Co-authorship spans first authorship on the Safety Nudges field study and a middle position on CoVeR, indicating steady team-based work across both research lines rather than solo output; the mid-range score reflects moderate rather than extensive network scale.
Activity40/100
Activity scores 40 on a sparse but recent record: counts by year show one work in 2024 and two in 2026, with 3 works over the past three years and no works appearing in the indexed 2025, suggesting a currently accelerating phase. The low score reflects record length rather than inactivity.
Reproducibility55/100
Reproducibility scores 55, the mean readiness across the 2 papers Evolve has reviewed, each rated medium. Safety Nudges reports a concrete 2-week field study (N=45) but lacks a control group and its auditor is unproven; CoVeR claims training-free, exact-budget token pruning with strong evidence at 8% retention, though significance reporting is weaker. Both are arXiv releases whose methods are checkable in principle.
Research summary
Varshini Elangovan is an early-stage researcher at Carnegie Mellon University whose profile is name-resolved from public bibliographic data (OpenAlex identity; no ORCID linked). The indexed record is young and small: three works over the past three years (one in 2024, two in 2026), with an h-index of 1 and an i10-index of 1. Two 2026 arXiv papers reviewed by Evolve split evenly across two fields — human-computer interaction (AI safety, 50% of the field distribution) and computer vision (efficient vision-language models, 50%). The first, "Safety Nudges: User-Facing Interventions for Real-Time AI Risk Awareness," is a first-author 2-week field study (N=45) of a browser extension that flags chatbot harms; it found awareness gains without significant behavior change, and Evolve notes the lack of a control group and an unproven auditor. The second, "CoVeR: Coverage-Based Token Pruning for Multi-View 3D Reasoning in VLMs," is a middle-author paper proposing training-free, coverage-based token pruning, claiming state-of-the-art results with exact budgets; Evolve judges the evidence strong at 8% retention, with weaker significance reporting. The record lists 14 distinct coauthors and 12 frequent collaborators, and both reviewed papers carry medium reproducibility readiness (mean 55). Overall this is a deliberately two-track early profile: real-time AI risk interfaces on one side, efficient 3D vision-language systems on the other, with citation impact not yet established.
Collaboration pattern
The record lists 14 distinct coauthors and 12 named frequent collaborators, including Sauvik Das, Virginia Smith, James Wedgwood, Chhavi Yadav, William Agnew, and Rakesh Ranjan. Both reviewed papers are team efforts: first authorship on the Safety Nudges field study and a middle position on CoVeR, consistent with collaboration across human-centered and vision/systems teams rather than a single-lab, solo-driven pattern.
Research keywords
Highlights
First-author field study of real-time AI safety nudges
Training-free token pruning for 3D vision-language models
A deliberately two-track early record
The bibliometric field split is exactly 50/50 between human-computer interaction (AI safety) and computer vision (efficient vision-language models), with five indexed topics each carrying a single work; both reviewed papers rate medium reproducibility readiness (mean 55).
Timeline
- 2024First indexed work appears in the OpenAlex record (1 publication in 2024), marking the start of a young bibliometric footprint.
- 2026Indexed output doubles to 2 works in a single year, split evenly across the profile's two fields.
- 2026First-author arXiv release: Safety Nudges — a 2-week field study (N=45) of a browser tool flagging real-time chatbot harms.
- 2026Middle-author arXiv release: CoVeR — training-free, coverage-based token pruning for multi-view 3D reasoning in vision-language models.
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.
- Profile is name-resolved from public bibliographic data (OpenAlex identity); no ORCID is linked, so identity confidence depends on the name-disambiguation step.
- Bibliometrics are OpenAlex-derived only; output and impact scores are marked missing for this young record, so no productivity or citation-impact estimate is offered.
- Only 2 papers have been reviewed by Evolve, both 2026 arXiv preprints; peer-reviewed venue acceptance is not captured in the provided data.
- Radar scores (collaboration 54, activity 40, focus 0, reproducibility 55) come from the provided bibliometric snapshot and may shift as the record grows.