Shiwali Mohan — preference learning & pluralistic alignment; first-authored work showing person-conditioned utility models beat universal ones
1
papers here
1
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-20) 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 in this snapshot. This axis would measure publication volume and cadence over time; with only one catalogued paper, no responsible score can be computed yet.
Impact—
Citation and downstream-influence data are absent here. This axis would track how often the work is cited and built upon by others; with a single recent arXiv paper, impact is not yet measurable.
Focus100/100
Field distribution is fully concentrated: the snapshot maps all reviewed work to a single field — machine learning (preference learning / pluralistic alignment / human-AI interaction) — producing the maximum focus score of 100. The one reviewed paper sits squarely in that field, though with a catalog of one, perfect concentration is easily attained and should be re-read as the catalog grows.
Collaboration36/100
The snapshot records 5 distinct co-authors, and the collaboration score of 36 sits in the lower-middle of the 0-100 scale. Recurring collaborators include Matt Hong, Dule Shu, Aniek Fransen, Shabnam Hakimi, and Matt Klenk. The reviewed paper is first-authored without a corresponding-author designation, so a small, stable core group is visible while overall network breadth remains limited in this early snapshot.
Activity—
No multi-year activity signals exist beyond the single 2026 entry. This axis would measure recent publishing tempo and ongoing research engagement; the snapshot is too thin to score it.
Reproducibility20/100
Reproducibility is the weakest measured axis: the mean readiness across 1 assessed paper is 20, and Evolve rates the reviewed paper's readiness as low. The verdict credits a convincing ablation but flags thin baselines, underspecified splits, and limited statistical reporting — a reporting-practice gap rather than a challenge to the reported results.
Research summary
Shiwali Mohan (ORCID 0000-0002-9299-4414) is a machine learning researcher working at the intersection of preference learning, pluralistic alignment, and human-AI interaction. The Evolve catalog currently holds one reviewed first-authored paper, "Learning Heterogeneous Preferences" (arXiv, 2026), which is the principal evidence base for this profile. The paper argues that utility models conditioned on the individual person outperform universal preference models and supports this at scale: person-conditioned models beat universal ones on 575k wheel-aesthetic judgments, with what Evolve's reviewers describe as a convincing ablation. The same review rates the paper's reproducibility readiness as low, citing thin baselines, underspecified splits, and limited statistical reporting — a reporting gap rather than a challenge to the result itself. Radar scores reflect a narrow, early snapshot: focus is maximal at 100, with all reviewed work mapped to a single field; collaboration scores 36 on the strength of 5 distinct co-authors, including recurring collaborators Matt Hong, Dule Shu, Aniek Fransen, Shabnam Hakimi, and Matt Klenk; reproducibility averages 20 across the one assessed paper. Output, impact, and activity axes lack sufficient data to score. Institution and country are not resolved in the current data, and this profile is name-resolved from public bibliographic records.
Collaboration pattern
The snapshot records 5 distinct co-authors, matching the five recurring collaborators listed: Matt Hong, Dule Shu, Aniek Fransen, Shabnam Hakimi, and Matt Klenk. With a collaboration score of 36, the network appears compact; the reviewed paper is first-authored without a corresponding-author flag, so the visible pattern is a small, stable core group around a first-author-led research agenda.
Research keywords
Highlights
Person-conditioned utility models outperform universal preference models
Reproducibility readiness rated low despite convincing ablation
Timeline
- 2026First-authored paper "Learning Heterogeneous Preferences" appears on arXiv, proposing person-conditioned utility models for subjective preference learning.
- 2026The paper's evaluation shows person-conditioned utility models beating universal ones on 575k wheel-aesthetic judgments, backed by a convincing ablation.
- 2026Evolve reviews the paper, rating reproducibility readiness as low due to thin baselines, underspecified splits, and limited statistical reporting.
Papers in the Evolve catalog
- first author
Learning Heterogeneous Preferences
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.
- Dimension scores come from a small bibliometric snapshot: 5 distinct co-authors, a single-field distribution, and only 1 paper with a reproducibility score.
- Output, impact, and activity axes are unscored because the underlying metrics are missing from this snapshot.
- Institution and country are not resolved in the provided data; the profile is name-resolved from public bibliographic records via ORCID 0000-0002-9299-4414.
- Substantive claims here rest on a single Evolve-reviewed paper; career-level generalizations should be treated as provisional.