Model-based RL researcher at Brown University: changepoint-aware world models and replay retention under dynamics shift
2
papers here
2
reviewed by Evolve
2
first-author
1
corresponding
—
works · OpenAlex
0
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—
Insufficient data to score: this snapshot records only 3 works (all 2026) with h-index and i10-index at 0, so no stable publication-volume picture exists yet. This axis would measure total research outputs across venues over the career.
Impact—
Insufficient data to score: citation counts and downstream adoption are not yet available for these 2026 arXiv preprints, and h-index is recorded as 0. This axis would measure citations, referencing works, and uptake by other research groups.
Focus0/100
The focus score of 0 reflects a dispersed OpenAlex topic record: five labeled topics — 3D Shape Modeling and Analysis (3), Generative Adversarial Networks and Image Synthesis (2), Surgical Simulation and Training (2), Robot Manipulation and Learning (1), and Interactive and Immersive Displays (1) — while the field distribution splits evenly, 0.5/0.5, between model-based RL/robot adaptation and robot learning/RL. The two reviewed papers themselves are tightly concentrated on replay and dynamics shift.
Collaboration28/100
The collaboration score of 28 rests on a very small recorded network: 3 distinct coauthors in the OpenAlex record, with Skye Thompson and George D. Konidaris listed as frequent collaborators. Both reviewed 2026 preprints are first-authored, one also corresponding, indicating an early, self-driven pattern rather than a large collaboration network.
Activity40/100
The activity score of 40 is computed from a thin recent record: recent3yWorks of 3, with countsByYear showing all 3 works in 2026 and no earlier years recorded. Both 2026 arXiv preprints appeared in the same year, so apparent output is concentrated in a single burst rather than sustained across years.
Reproducibility55/100
Reproducibility scores 55 (reproMean across 2 repro-scored papers, both rated medium readiness by Evolve). Review verdicts flag concrete gaps: one task with three seeds and no statistical tests in the changepoint-aware world models paper, and five seeds with self-tuned thresholds limiting generality in the replay retention study.
Research summary
Everest Yang is an early-career researcher at Brown University working on model-based reinforcement learning, focused on how agents handle non-stationary dynamics. In 2026 two first-authored arXiv preprints appeared: 'Changepoint-Aware World Models', in which CUSUM-triggered replay flushing in DreamerV3 recovers faster than respawn and passive post-shift strategies — though the evidence spans one task, three seeds, and no statistical tests — and 'Characterizing Replay Retention Under Dynamics Shift', which argues that replay value depends on shift magnitude and recurrence rather than age alone, with well-controlled but seed-limited evidence (five seeds, self-tuned thresholds). Both papers are first-authored; the changepoint work is also corresponding-author. The OpenAlex record shows 3 works in the last three years (all 2026), 3 distinct coauthors — with Skye Thompson and George D. Konidaris as frequent collaborators — and an even field split (0.5/0.5) between machine learning (model-based RL / robot adaptation) and robot learning / reinforcement learning. Reproducibility averages 55 across the 2 scored papers, both rated medium. h-index and i10-index are recorded as 0, so output and impact axes cannot yet be scored. OpenAlex topic labels (3D shape modeling, GAN-based image synthesis, surgical simulation) diverge from the reviewed RL papers, suggesting topic or identity noise in the record. This profile is name-resolved from public bibliographic data.
Collaboration pattern
Everest Yang's recorded network is small: OpenAlex lists 3 distinct coauthors, and the frequent-collaborator set comprises Skye Thompson and George D. Konidaris (the researcher's own name also appears in that list, likely a record artifact). Both 2026 preprints are first-authored, one also corresponding, consistent with an early-career, self-driven pattern; the collaboration score is 28 of 100. Per-paper coauthor roles are not in the snapshot.
Research keywords
Highlights
CUSUM-triggered replay flushing speeds recovery in DreamerV3
Replay value tracks shift magnitude and recurrence, not age
Timeline
- 2026First-authored arXiv preprint 'Characterizing Replay Retention Under Dynamics Shift in Model-Based Reinforcement Learning' appears, arguing replay value depends on shift magnitude and recurrence; Evolve rates its reproducibility readiness medium.
- 2026First- and corresponding-authored preprint 'Changepoint-Aware World Models' appears on arXiv, applying CUSUM-triggered replay flushing in DreamerV3 for faster recovery from dynamics shifts.
- 2026Year-end OpenAlex record shows all 3 recent works fall in 2026, with 2 papers scored for reproducibility (mean 55) and both first-authored preprints cataloged and reviewed by Evolve.
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.
- All bibliometrics come from a single OpenAlex record with no ORCID linked; 'Everest Yang' may not be uniquely resolvable by name, so identity confidence is limited.
- Coverage is limited to the two 2026 arXiv preprints reviewed by Evolve; no peer-reviewed venue record or citation data is available in this snapshot.
- h-index and i10-index are recorded as 0 and countsByYear covers only 2026, so the output and impact axes cannot be scored and no career-level trend exists yet.
- OpenAlex topic labels (3D shape modeling, surgical simulation, GAN image synthesis) diverge from the reviewed papers' model-based RL focus, and the frequent-coauthor list repeats the researcher's own name — both signs of topic/identity noise in the record.