NLP researcher focused on diffusion language models, GraphRAG retrieval, and process-level supervision for reasoning
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
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 yet: the underlying data lacks a verified publication count, venue breakdown, and per-year output for this name. This axis would measure total papers, first/corresponding-author share, and publication venues once a resolved author record is available.
Impact—
No impact score yet: citation counts, h-index, and per-paper influence are absent from the snapshot, and neither cataloged paper has citation data. This axis would measure citations, citing works, and downstream adoption once citation records are linked.
Focus100/100
The focus score is 100, the maximum: the field distribution for this author records exactly one field, machine learning / natural language processing, with no dispersion into adjacent areas. Both cataloged 2026 papers sit inside this single lane — diffusion-based reasoning and process supervision for grounded generation — suggesting deliberate specialization, though a single-field count on a short record can also reflect limited coverage.
Collaboration56/100
The collaboration score is 56, built from 16 distinct coauthors in the snapshot. The frequent-coauthor list is dominated by a repeating cluster — Senlei Zhang, Qian-Wen Zhang, and Junnan Dong each appear more than once — alongside Linhao Luo, Siyu An, Shuhao Zhang, Junjie Huang, and Jiarui Qin, indicating a stable core team rather than many one-off ties. Author positions (last on LADDER, middle on the other paper) suggest both supervisory and contributing roles.
Activity—
No activity score yet: the snapshot has no per-year publication series, so recent-year output cadence cannot be computed. This axis would measure publication frequency and momentum across years; only two 2026 arXiv entries exist here, too thin to score.
Reproducibility55/100
The reproducibility score is 55, computed as the mean over a single assessed paper (reproPapers = 1), so it should be read as a weak signal. That paper is LADDER, whose readiness is rated medium; Evolve's verdict reports a claimed +5.6 EM gain and 4.1x speedup but notes the gains lack significance tests — exactly the kind of gap that caps reproducibility confidence until code, seeds, and statistical rigor are verified.
Research summary
Xing Sun is an early-stage natural language processing researcher whose visible record centers on efficient, trustworthy reasoning for large language models. The Evolve catalog holds two 2026 arXiv papers. On 'LADDER: Graph-Guided Diffusion Language Models for Efficient Multi-Hop Reasoning' Sun is the last author — a slot usually reserved for senior or supervising contributors; Evolve's review reports a claimed +5.6 exact-match gain and a 4.1x speedup from firing GraphRAG retrieval as entities emerge during diffusion decoding, while flagging the absence of significance tests. The second paper, 'From Final Artifacts to Trajectories', places Sun in a middle author position and shifts supervision from final answers to intermediate reasoning trajectories for evidence-grounded long-form generation. Bibliometrically the profile is unusually concentrated: the field distribution records exactly one field (machine learning / natural language processing), producing a maximum focus score of 100. Collaboration is moderate — 16 distinct coauthors yield a score of 56 — anchored by a repeating core of Senlei Zhang, Qian-Wen Zhang, and Junnan Dong. Reproducibility stands at 55 from a single assessed paper (medium readiness). Output, impact, and activity remain unscored for lack of publication-count and citation data, and no institution, country, or ORCID is recorded; identity is name-resolved from public bibliographic data.
Collaboration pattern
Xing Sun works within a compact, recurring team: of 16 distinct coauthors, Senlei Zhang, Qian-Wen Zhang, and Junnan Dong each appear repeatedly in the frequent-coauthor list, alongside Linhao Luo, Siyu An, Shuhao Zhang, Junjie Huang, and Jiarui Qin. Taking the last-author slot on LADDER and a middle slot on the trajectory-supervision paper suggests a mix of senior supervision and hands-on contribution. The collaboration score is 56; no institutional affiliation is recorded to contextualize these ties.
Research keywords
Highlights
Graph-guided diffusion decoding meets GraphRAG for multi-hop reasoning
Shifting evaluation from final answers to reasoning trajectories
The second cataloged 2026 arXiv paper, 'From Final Artifacts to Trajectories: Retrospective Process Supervision for Evidence-Grounded Long-Form Generation', lists Sun as a middle author. It proposes supervising the generation process retrospectively rather than judging only final artifacts, targeting evidence-grounded long-form output; Evolve has not yet published a verdict, keywords, or reproducibility rating for it.
A two-pronged 2026 agenda: efficient retrieval-grounded reasoning plus process-level supervision
Taken together, Sun's two 2026 entries outline one agenda: LADDER speeds up multi-hop reasoning with graph-guided diffusion (claimed +5.6 EM, 4.1x speedup), while the trajectory-supervision paper improves how long-form, evidence-grounded output is supervised — both inside the single field his record shows (machine learning / natural language processing).
Timeline
- 2026Publishes LADDER (arXiv), a graph-guided diffusion language model for efficient multi-hop reasoning, in the last author position.
- 2026Co-authors 'From Final Artifacts to Trajectories', introducing retrospective process supervision for evidence-grounded long-form generation (middle author).
- 2026Bibliometric snapshot places both papers in the machine learning / natural language processing field, yielding a maximum focus score of 100.
- 2026Evolve reviews LADDER, rating its reproducibility readiness medium; the reproducibility mean is 55 across 1 assessed paper.
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 from OpenAlex-derived data only; no citation counts, h-index, or venue-quality metrics are available, so output, impact, and activity axes are unscored.
- 'Xing Sun' is a common name; this profile is name-resolved from public bibliographic data, and institution, country, and ORCID are unrecorded.
- Only two papers (both 2026, arXiv) are in the reviewed catalog; the reproducibility score of 55 rests on a single assessed paper.
- The second cataloged paper has no Evolve verdict, keywords, field tag, or reproducibility rating yet.