Xinyu Zhang
—
Researcher spanning LLM agent self-evolution and long-context speech benchmarking; co-author on two 2026 arXiv papers
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-12) 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 measures publication volume and productivity over time. The underlying counts are unavailable in this snapshot — beyond the two Evolve-reviewed arXiv preprints, no totals or yearly productivity data exist — so the axis is left unscored.
Impact—
Impact tracks citations, uptake, and influence of the work. No citation counts or derived impact metrics are available for this researcher yet, so the axis cannot be scored; it would measure how widely the two preprints are adopted.
Focus0/100
The focus score is 0, reflecting an even 50/50 split in the field distribution between machine learning work on LLM agent self-evolution and speech and natural language processing. Rather than a deep single-track agenda, the two catalogued 2026 papers sit in distinct subfields — one an agent memory framework, the other a spoken-language benchmark — so concentration across fields is minimal by this metric.
Collaboration54/100
Collaboration scores 54, based on 14 distinct coauthors in the snapshot. Both papers carry middle, non-corresponding authorship, consistent with contribution inside larger teams rather than sole leadership. The recurring collaborator list includes Ruimin Pei, Yongkang Wu, Shangyi Zheng, Deyang Li, Xize Cheng, and Dongjie Fu.
Activity—
Activity measures recent publishing cadence — papers per year or per quarter. The snapshot provides no publication-frequency data, and both catalogued papers carry the same 2026 year, so a cadence cannot be established; the axis is left unscored rather than inferred.
Reproducibility55/100
Reproducibility averages 55 across the 2 papers Evolve has reviewed, and both are rated medium readiness. The reviews flag concrete limits: TTSE's pivotal ablation lacks significance and its warehouse component drives most of the SOPBench gain, while Vox-Infinity relies on LLM-built QA, LLM judging, and only partial data release — leaving room to strengthen artifacts and statistics.
Research summary
Xinyu Zhang is an early-stage, name-resolved profile in the Evolve catalog, currently represented by two 2026 arXiv preprints. The field distribution splits evenly at 50/50 between machine learning work on LLM agent self-evolution and speech and natural language processing, so the profile bridges two adjacent AI areas rather than one narrow track. The first paper, TTSE, proposes a two-track online self-evolution framework that separates declarative and procedural experience memory (FACT/TIP) under an explicit rule lifecycle and risk theory; Evolve's review finds the risk framing clean but notes that a pivotal ablation lacks significance and that the warehouse component drives most of the SOPBench gain. The second, Vox-Infinity, claims the first provenance-annotated long-context benchmark for spoken language models and documents a solid recency effect, though LLM-built QA, LLM judging, and partial data release limit rigor. Across the snapshot Zhang records 14 distinct coauthors and a moderate collaboration score of 54, holding middle, non-corresponding authorship on both papers — a contributor role within larger teams rather than sole leadership. Reproducibility averages 55 across the two reviewed papers, both rated medium readiness, consistent with the statistical and release caveats above. Output, impact, and activity axes lack sufficient data and are deliberately left unscored. No institution, country, or ORCID is available in this snapshot.
Collaboration pattern
Xinyu Zhang's snapshot records 14 distinct coauthors and a moderate collaboration score of 54. Both reviewed papers are middle-author, non-corresponding contributions, suggesting work embedded in larger teams. Recurring names include Ruimin Pei, Yongkang Wu, Shangyi Zheng, Yaqing Zhang, Deyang Li, Xiang Zhang, Xize Cheng, Wenxu Jia, Chenyuhao Wen, and Dongjie Fu. No first or corresponding role appears in the reviewed sample, so independent leadership is not yet evidenced.
Research keywords
Highlights
TTSE: a dual-track self-evolution framework for LLM agents
Vox-Infinity: first provenance-annotated long-context speech benchmark
Timeline
- 2026TTSE, a two-track online self-evolution framework for LLM agents, appears on arXiv in the machine learning (LLM agent self-evolution) field, with middle authorship.
- 2026Vox-Infinity, a benchmark probing the limits of long-context spoken language models, appears on arXiv in speech and natural language processing, also with middle authorship.
- 2026Both papers enter the Evolve catalog and are reviewed; reproducibility averages 55 across the 2 papers, each rated medium 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.
- Bibliometric dimensions are computed from OpenAlex-derived bibliometric data only; the output, impact, and activity axes lack sufficient data and are not scored.
- "Xinyu Zhang" is a common name; this profile is name-resolved from public bibliographic data and may not be guaranteed to correspond to a single individual.
- No institution, country, or ORCID is available, so affiliation and career context are unknown.
- The reproducibility score rests on only 2 Evolve-reviewed arXiv preprints (mean 55, both medium readiness) and may not reflect peer-reviewed output.