First author of GraphEcho: benchmarking retrieval redundancy and evidence provenance in LLM graph agents
1
papers here
1
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-21) 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—
Not yet scored: the current data contains no publication counts, venue list, or output trajectory for this researcher. This axis would measure the volume, venues, and cadence of published work over time.
Impact—
Not yet scored: citation counts, h-index and downstream adoption are absent from the input data. This axis would measure how widely the GraphEcho benchmark and related outputs are cited and reused.
Focus100/100
A focus score of 100 reflects the current field distribution: cataloged output maps entirely to one field, artificial intelligence (LLM agents / retrieval-augmented generation). With a single reviewed paper, specialization is undiluted, though the score should be read as early-stage concentration rather than an established, long-run research agenda.
Collaboration28/100
The collaboration score of 28 reflects a compact working pattern: the cataloged record lists 3 distinct coauthors, with Yixi Zhou, Lei Fan and Fan Zhang recurring. The one reviewed paper is first-authored without a corresponding-author role, consistent with a small-team style rather than large multi-instruction consortia. Broader network metrics are not yet available.
Activity—
Not yet scored: only one 2026 record is in scope, so recent-year publication cadence cannot be assessed. This axis would measure ongoing output and research engagement across recent years.
Reproducibility55/100
Reproducibility averages 55 across 1 assessed paper. Evolve judged GraphEcho's reproducibility readiness as medium: the benchmark offers controlled conditions and rigorous statistics, but reliance on a single training seed and a narrow set of models tempers how far conclusions generalize. Reproducers should budget for seed-level re-runs.
Research summary
Sikun Wang is an early-stage researcher in artificial intelligence whose cataloged work centers on evaluation and reliability of LLM-based graph agents. The Evolve catalog contains one first-authored paper, 'GraphEcho: Structural Redundancy and Evidence Provenance in LLM Graph Agents' (arXiv, 2026), which introduces a controlled benchmark separating path redundancy in retrieved evidence from evidence provenance — asking whether graph-agent failures stem from redundant retrieval or from untracked source attribution. Evolve's review credits the work with rigorous statistical treatment while noting that a single training seed and a narrow set of evaluated models temper the conclusions; reproducibility readiness is rated medium, with a mean score of 55 across one assessed paper. The field profile is highly concentrated: all cataloged output maps to a single field, artificial intelligence (LLM agents / retrieval-augmented generation), yielding a focus score of 100. Collaboration metrics indicate a compact pattern — 3 distinct coauthors, with Yixi Zhou, Lei Fan and Fan Zhang recurring — and a collaboration score of 28. Output, impact and activity dimensions are not scored because the underlying bibliometric data is missing; no institution, country or ORCID was available, so this profile is name-resolved from public bibliographic data and should be read as an early-stage view rather than a complete record.
Collaboration pattern
The available record shows a small, stable team: 3 distinct coauthors in total, with Yixi Zhou, Lei Fan and Fan Zhang appearing as frequent collaborators on the reviewed work. Sikun Wang led as first author (non-corresponding) on GraphEcho, suggesting a hands-on experimental role within a compact group rather than a large distributed consortium. Co-author roles beyond the cataloged paper are not yet visible in the data.
Research keywords
Highlights
GraphEcho: a controlled benchmark for LLM graph agents
Rigorous statistics, honestly bounded
Training meets measurement: GRPO and source counting
Timeline
- 2026First-authored preprint 'GraphEcho: Structural Redundancy and Evidence Provenance in LLM Graph Agents' released on arXiv.
- 2026Evolve reviewed the paper into its catalog, crediting a controlled benchmark that separates path redundancy from provenance with rigorous statistics.
- 2026Reproducibility readiness assessed as medium, with a mean reproducibility score of 55 over 1 paper; single-seed training and narrow model coverage noted.
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.
- Profile is name-resolved from public bibliographic data; no ORCID, institution, or country was available to confirm identity, and the name may not be unique.
- All scores derive from a single cataloged paper (reproPapers = 1); treat radar dimensions as early-stage signals rather than established track-record metrics.
- Output, impact and activity axes are unscored because the underlying bibliometric data is missing, not because the underlying values are zero.
- Catalog coverage is limited to one arXiv preprint; conference or journal publications outside this snapshot are not reflected.