RL researcher at Apple focused on reset-free RL evaluation and skill-library learning for robot manipulation.
2
papers here
1
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
1
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 output score is available yet: this axis would measure publication volume and productivity over time, but the indexed record (five works spanning 2022-2026) is too sparse and name-resolution-dependent to score fairly.
Impact—
No impact score is available yet: this axis would measure citation influence (citation counts, h-index trajectory, i10-index). The recorded h-index of 1 and i10-index of 0 leave too little signal to score meaningfully.
Focus100/100
Focus scores 100/100: the field distribution records a single field — reinforcement learning / robotics — and OpenAlex topics concentrate on Reinforcement Learning in Robotics, Robot Manipulation and Learning, and Adversarial Robustness in Machine Learning (two papers each), with smaller residues in hate-speech detection and topic modeling. Both 2026 catalog preprints sit squarely in robot learning, indicating a tightly concentrated research agenda.
Collaboration14/100
Collaboration scores 14/100, reflecting a very small recorded footprint: exactly one distinct coauthor, with Chandru Venkataraman the sole recurring partner in the catalog. Both reviewed 2026 preprints are first-authored and non-corresponding. The sample is small, so the score should be read as 'little data' as much as 'few collaborators', and may shift quickly as more works are indexed.
Activity47/100
Activity scores 47/100. OpenAlex counts one work in 2022, two in 2025 and two in 2026 — four works within the last three years. The 2026 output is concentrated: two first-authored arXiv preprints in robot learning appeared in the same year. The mid-range score reflects the 2022-2025 gap rather than current output, which appears to be accelerating.
Reproducibility55/100
Reproducibility scores 55/100, computed from one reviewed paper (reproPapers: 1, mean 55). Evolve rates both catalog preprints 'medium' reproducibility readiness: REVERSAL-BENCH is simulation-only with thin quantitative reporting, and REFACTOR-VLA's headline LIBERO result depends on task-label supervision. The evidence base is a single paper, so treat this as an early signal rather than a stable score.
Research summary
Riyaaz Shaik is a reinforcement learning researcher at Apple in the United States. This profile is name-resolved from public OpenAlex bibliographic data, which attributes five works to this identity — one in 2022, two in 2025 and two in 2026, four of them within the last three years — with a recorded h-index of 1 and i10-index of 0, marking an early-stage record. Evolve's catalog carries two 2026 arXiv preprints with Shaik as first author. REVERSAL-BENCH introduces a reversibility axis (rho) and a simulator-based reset oracle across five physics engines, documenting a sharp reset-free RL 'cliff'; Evolve judges the design well made and causally controlled but simulation-only, with thin quantitative reporting. REFACTOR-VLA learns typed motor-program skill libraries through a wake/sleep loop and beats baselines 4/4 on LIBERO NMI, though the headline result relies on task-label supervision, straining its 'unsupervised' framing. OpenAlex topics concentrate on Reinforcement Learning in Robotics, Robot Manipulation and Learning, and Adversarial Robustness, with residue from earlier-topic areas such as hate-speech detection and topic modeling. The recorded collaboration footprint is narrow — one distinct coauthor, with Chandru Venkataraman the recurring partner — and citation-impact data remain too sparse to score; reproducibility review of one paper yields a mean readiness of 55/100.
Collaboration pattern
The recorded collaboration footprint is very small: OpenAlex lists exactly one distinct coauthor, and the collaboration dimension scores 14/100. Chandru Venkataraman is the sole recurring partner in the catalog. Both reviewed 2026 preprints are first-authored and non-corresponding, consistent with a small-team or solo-lead working style. Because the indexed sample is thin, this pattern describes the data as much as the habit, and should be revisited as more works are attributed.
Research keywords
Highlights
REVERSAL-BENCH: quantifying the reset-free RL cliff
REFACTOR-VLA: wake/sleep skill libraries beat baselines on LIBERO
A 2026 two-paper arc: from irreversibility benchmark to skill-library method
Two same-year first-authored preprints form a complementary pair — a benchmark that measures irreversibility (REVERSAL-BENCH) and a skill-library learning method (REFACTOR-VLA) — sketching a robotics-RL agenda built on reversibility and skill abstraction; both are rated medium reproducibility readiness.
Timeline
- 2022Earliest indexed activity for this identity in the OpenAlex snapshot: one publication recorded in 2022.
- 2025Two works recorded in 2025, resuming output after the 2022 entry in the bibliographic snapshot.
- 2026Most active year on record: two works, including both first-authored arXiv preprints in the Evolve catalog — REVERSAL-BENCH and REFACTOR-VLA, both in robot learning.
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.
- Bibliometrics come from OpenAlex only (identity: openalex); no ORCID was provided, so name resolution may miss or merge works.
- Output and impact dimensions are unscored: the record (five works, h-index 1, i10-index 0) is too sparse for a fair estimate.
- Reproducibility (55/100) is computed from a single reviewed paper, so it is an early signal rather than a stable score.
- Evolve review coverage covers only the two 2026 arXiv preprints; other OpenAlex-indexed works are unreviewed.