Recommender-systems researcher advancing generative recommendation with semantic IDs and reinforcement learning
2
papers here
0
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-27) 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 publication-count or output metrics are available yet: the underlying data covers only the two reviewed 2026 preprints. This axis would measure total output volume, venue distribution, and publication cadence once a fuller bibliographic record exists.
Impact—
Citation counts, h-index, and downstream influence are absent from the current bibliometric snapshot, so no impact score can be reported. This axis would measure citations to the reviewed papers and their uptake in generative-recommendation research.
Focus—
Topic-focus metrics are not yet available in the snapshot. The two reviewed papers suggest a concentration on semantic-ID-based generative recommendation, but this remains a qualitative reading; this axis would quantify topical concentration across the full record.
Collaboration41/100
The collaboration dimension scores 41 on a 0-100 scale, computed from real bibliometric data: the snapshot records 7 distinct coauthors, including Mengdan Zhu, Yufan Zhao, Sophie Di, Yao Zhao, Tao Di, Sridhar Iyer, and Liang Zhao. Both cataloged papers list Yan in a middle authorship position and not as corresponding author, suggesting participation within a larger team rather than leading an independent group — consistent with a moderate, team-embedded collaboration pattern.
Activity—
Recent-activity metrics such as annual publication counts or year-over-year trends are unavailable; the snapshot pins only two 2026 preprints. This axis would measure publishing cadence and sustained activity once fuller bibliographic records exist.
Reproducibility—
The aggregate reproducibility axis is not yet scored — no code-availability or artifact data underlies it. Qualitatively, both reviewed papers are tagged medium reproducibility readiness; this axis would quantify artifact release and replication evidence across the full record.
Research summary
Yulan Yan is a researcher in recommender systems and information retrieval whose cataloged work centers on generative recommendation built on semantic IDs (SID). The Evolve catalog holds two 2026 arXiv preprints, in both of which Yan is a middle author and not the corresponding author. The first, 'From Interests to Semantic IDs', develops retrieval-grounded credit assignment using reinforcement learning with group-relative policy optimization (GRPO) and process supervision; Evolve's review credits it with improving SID-based recommendation across three datasets and well-designed ablations, while noting the significance is partial and reward leakage untested. The second, 'Learning Better Reasoning for Generative Recommendation with Semantic IDs' (Evo-Rec), combines best-of-n reasoning supervised fine-tuning with NDCG-reward GRPO; the review reports gains on three Amazon datasets with solid ablations, but flags missing seed variance and second-hand baselines. Both papers are tagged medium reproducibility readiness. The bibliometric snapshot records 7 distinct coauthors — among them Mengdan Zhu, Yufan Zhao, and Sridhar Iyer — and a collaboration dimension score of 41, consistent with team-embedded work on reinforcement learning for generative recommendation. Institution, country, and ORCID are unavailable, so this profile is name-resolved from public bibliographic records.
Collaboration pattern
The snapshot records 7 distinct coauthors: Mengdan Zhu, Yufan Zhao, Sophie Di, Yao Zhao, Tao Di, Sridhar Iyer, and Liang Zhao, giving a collaboration dimension score of 41 (0-100). On both cataloged 2026 preprints Yan is a middle author and not corresponding author, pointing to contribution within a multi-author team rather than sole or senior leadership. The frequent co-author set suggests an ongoing group working on reinforcement learning for generative recommendation.
Research keywords
Highlights
Retrieval-grounded credit assignment for SID generative recommendation
Evo-Rec: reasoning-enhanced generative recommendation with semantic IDs
A coherent GRPO-centered research thread
Both 2026 preprints share one recipe — semantic IDs as representation, GRPO-style reinforcement learning as optimizer, and reward design as the lever (process supervision in one, NDCG reward in the other) — suggesting an early but focused research agenda.
Timeline
- 2026arXiv preprint 'From Interests to Semantic IDs: Retrieval-Grounded Credit Assignment for Generative Recommendation' appears; Evolve reviews it as improving SID recommendation on three datasets with good ablations, though significance is partial and reward leakage untested.
- 2026arXiv preprint 'Learning Better Reasoning for Generative Recommendation with Semantic IDs' (Evo-Rec) appears, combining best-of-n reasoning SFT with NDCG-reward GRPO; reviewed as lifting SID generative recommendation on three Amazon datasets.
- 2026Both preprints enter the Evolve catalog with medium reproducibility readiness; the bibliometric snapshot records 7 distinct coauthors and a collaboration dimension score of 41.
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 come from a bibliometric snapshot only; five of six dimensions (output, impact, focus, activity, reproducibility) lack sufficient data and remain unscored.
- Institution, country, and ORCID are missing; the profile is name-resolved from public bibliographic data and may merge or split namesakes.
- Evidence rests on two 2026 arXiv preprints reviewed by Evolve, both middle-author and non-corresponding with medium reproducibility readiness, so conclusions about independent research leadership are limited.
- Timeline granularity is limited to a single year (2026); earlier career history and peer-reviewed venue acceptance status are not available in the input data.