Machine learning researcher at Dartmouth studying representational geometry and world-model composition in LLM agents
2
papers here
1
reviewed by Evolve
0
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 score yet: this snapshot carries no publication counts, venues, or multi-year output totals, so productivity cannot be estimated. This axis would measure total research output — paper volume, venues, and author roles — accumulated over time.
Impact—
No score yet: citation counts, h-index, and downstream adoption are absent from the available data, so influence cannot be measured. This axis would capture how widely the work is cited and built upon by other researchers.
Focus100/100
Scores 100. The only labeled field in the distribution is machine learning (representation learning), attached to the reviewed displacement-geometry paper. The second 2026 preprint, on world-model and policy composition in LLM agents, is not yet field-labeled, so within the labeled evidence the agenda is a single concentrated field rather than a spread; the maximal score reflects that narrow, consistent labeling on a small sample.
Collaboration44/100
Scores 44, a moderate collaboration level. The snapshot records 8 distinct coauthors, with Chenming Shang, Ruize Xu, Xiao Yu, Nikhil Singh, Jun Jie Ou Yang, and Adam Breuer most frequent; Ruize Xu and Nikhil Singh each appear twice. On both 2026 preprints Tang sits in a middle author position without corresponding status, fitting a contributing member of small research teams rather than a solo author or a large-consortium hub.
Activity—
No score yet: per-year publication frequency and recent-year output are not in the snapshot, so research momentum cannot be assessed. This axis would measure publishing cadence and the continuity of output across recent years.
Reproducibility100/100
Scores 100, on a narrow base: the mean is computed from exactly 1 reviewed paper (reproPapers: 1), the displacement-geometry study, which Evolve rates high in reproducibility readiness. The perfect score should be read as 'the single assessable paper is straightforward to rerun,' not as an established track record; additional reviewed papers would be needed to confirm the pattern.
Research summary
Yujin Tang is a machine learning researcher affiliated with Dartmouth College whose cataloged work centers on representation learning: how models organize concepts and how learned structure can be compared across systems. The Evolve catalog lists two 2026 arXiv preprints, both with Tang in a middle author position and neither as corresponding author. The first, 'Displacement Geometry Captures Platonic Shared Reality Across Models and Modalities,' argues that models share displacement directions rather than positions, drawing on 44 encoders and reframing the debate around the Platonic Representation Hypothesis; Evolve finds the argument compelling while flagging whitening and data-overlap choices for scrutiny, and rates its reproducibility readiness high. The second, 'How do World Models and Policies Compose in LLM Agents? A Joint Spectral and Behavioral Account,' turns to a neighboring question about world-model and policy composition in LLM agents but is not yet field-labeled or reviewed. Radar scores are sharply polarized: focus 100 (one labeled field, machine learning/representation learning), reproducibility 100 computed from a single reviewed paper, and collaboration 44, with 8 distinct coauthors including Chenming Shang, Ruize Xu, Xiao Yu, Nikhil Singh, Jun Jie Ou Yang, and Adam Breuer. Output, impact, and activity lack underlying data and are left unscored.
Collaboration pattern
Tang works within small, recurring teams: the snapshot lists 8 distinct coauthors, with Chenming Shang, Ruize Xu, Xiao Yu, Nikhil Singh, Jun Jie Ou Yang, and Adam Breuer appearing most often (Ruize Xu and Nikhil Singh each twice). On both 2026 arXiv preprints Tang holds a middle author position without corresponding-author status, consistent with a hands-on contributor inside a lab-style group rather than a solo author or consortium lead.
Research keywords
Highlights
Displacement geometry, not position, as the shared substrate across 44 encoders
Joint spectral and behavioral account of world-model and policy composition in LLM agents
Tang co-authored this 2026 arXiv preprint in a middle author position; the catalog entry carries no field label, keywords, verdict, or reproducibility rating yet, so this profile treats it as an early, unreviewed extension of the representation-geometry agenda.
High reproducibility readiness on the sole reviewed paper
Timeline
- 2026Co-authored (middle position) 'Displacement Geometry Captures Platonic Shared Reality Across Models and Modalities' on arXiv.
- 2026Co-authored (middle position) 'How do World Models and Policies Compose in LLM Agents? A Joint Spectral and Behavioral Account' on arXiv.
- 2026The displacement-geometry paper entered the Evolve catalog with a reviewed verdict and high reproducibility readiness.
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 are computed from a public bibliometric snapshot only; no citation counts, h-index, or venue-level metrics were available.
- The reproducibility score of 100 rests on a single reviewed paper (reproPapers: 1) and should not be read as an established track record.
- The second 2026 preprint has no field label, keywords, verdict, or slug in the catalog; it is listed but not yet reviewed.
- No ORCID or country is available, and 'Yujin Tang' is a common-style name; this profile is name-resolved from public bibliographic data linked to Dartmouth College.