Meta researcher spanning recommender systems, self-evolving LLM research agents, and multi-agent workflow engineering
4
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
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—
Output cannot be scored: the OpenAlex snapshot lacks reliable publication aggregates for this profile, and the reported h- and i10-indexes read zero, treated here as missing rather than true zeros. This axis would measure publication volume, venue tier, and authorship share.
Impact—
Impact is not scored: no citation counts or uptake signals are available in the current snapshot, so influence cannot be estimated. This axis would measure citations, benchmark adoption, and downstream reuse of the 2026 preprints.
Focus0/100
The focus score of 0 reflects maximal spread rather than thin coverage: the field distribution splits evenly at about one third each across recommender systems/ML, autonomous LLM research agents, and LLM-based multi-agent systems. Topic counts likewise span recommender systems (8), bandit algorithms (8), information retrieval (7), topic modeling (6), and NLP (5). The 2026 catalog papers still form one coherent thread around self-evolving, LLM-driven systems.
Collaboration74/100
A collaboration score of 74 rests on 41 distinct coauthors in the recent window and a stable recurring group of twelve, including Hanqing Zeng, Ming Li, Dai Li, Xuying Ning, Bo Sun, and Rui Li. All four catalogued 2026 papers list the researcher in a middle authorship position and none as corresponding author, indicating a contributor role inside larger Meta teams.
Activity81/100
Activity scores 81 on the strength of 15 works within the recent three-year window, with the year-level count attributing 15 works to 2026 alone. All four papers in the Evolve catalog appeared as 2026 arXiv preprints across recommender systems, autonomous research agents, and multi-agent LLM workflows, indicating a dense, currently active publishing cadence concentrated in the present year.
Reproducibility43/100
Reproducibility averages 43 across the three Evolve-reviewed papers, whose readiness ratings are medium, low, and medium. The verdicts flag concrete gaps: LION's theoretical proofs are thin, Auto-RecSys reports operational but no model-quality evidence, and the graph-engineering paper lacks significance testing with GSM8K tempering its claims. Independent replication of these results would likely require substantial effort.
Research summary
Yinglong Xia is a researcher at Meta (United States), with this profile name-resolved from OpenAlex bibliographic data linked to ORCID 0000-0002-8155-5440. The snapshot's field distribution splits evenly into three threads: recommender systems and machine learning, autonomous LLM research agents for ML systems, and LLM-based multi-agent systems. The 2026 Evolve catalog reflects this spread: a sparse-memory method for continual generative recommendation, Meta's Auto-RecSys harness in which self-evolving LLM agents run industry-scale recommender experiments, a training-free, inference-time graph-compilation method for multi-agent workflows, alongside a study of knowledge distillation during mid-training. Topic counts — recommender systems (8), advanced bandit algorithms (8), information retrieval and search behavior (7), topic modeling (6), and natural language processing techniques (5) — frame a recommendation-centered core now extending into agentic systems. Activity is high, with 15 works in the recent three-year window and 41 distinct coauthors, anchored by a recurring group of twelve. All catalogued papers are 2026 arXiv preprints in middle authorship positions, none corresponding, and reviewed reproducibility readiness is medium to low. Output and impact metrics are unavailable in the current snapshot and are treated as missing rather than scored.
Collaboration pattern
Collaboration is broad but team-based: 41 distinct coauthors appear in the recent window, with a recurring core of twelve including Hanqing Zeng, Ming Li, Dai Li, Xuying Ning, Bo Sun, Rui Li, Yi Zhang, Silvia Gong, Xuan Cao, Cornelia Carapcea, Qunshu Zhang, and Zhigang Wang. All four catalogued 2026 papers carry middle authorship and no corresponding role, consistent with contributing within larger Meta project teams.
Research keywords
Highlights
Sparse KV memory tames evolution conflict in generative recommendation
Autonomous research agents run industry-scale RecSys experiments
Training-free graph compilation for multi-agent LLM workflows
Timeline
- 2026Self-Evolving Memory for Generative Recommendation posted to arXiv, introducing LION with a sparse key-value memory and consolidation loss for continual generative recommendation.
- 2026Auto-RecSys preprint describes autonomous LLM research agents running industry-scale recommender-system experiments with self-evolving playbooks at Meta.
- 2026Inference-Time Graph Engineering for Multi-Agent LLM Workflows introduces training-free one-shot temporal graph compilation, improving 5 of 6 benchmarks.
- 2026Knowledge Distillation During Mid-Training Favors Reasoning over Factual Recall enters the catalog; it has not yet been 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.
- Bibliometrics are computed from an OpenAlex snapshot; citation-based output and impact metrics are absent (reported indexes read zero) and are treated as missing, not as true zeros.
- Identity is name-resolved from public bibliographic data linked to ORCID 0000-0002-8155-5440; the name Yinglong Xia appears across multiple records, so affiliation matching carries residual ambiguity.
- All four catalogued works are 2026 arXiv preprints with middle authorship and no corresponding role; Evolve has reviewed three of the four, and the mid-training knowledge-distillation paper has no verdict yet.
- Year-level publication counts cover only 2026, so multi-year productivity trends cannot be established from this snapshot.