Fudan University NLP researcher building audited benchmarks for scientific reasoning and peer-review evaluation
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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 yet be scored: the OpenAlex snapshot contains no cumulative publication volume and reports an h-index and i10-index of 0, so this axis — which would measure total scholarly production and its trajectory — is left unscored rather than estimated.
Impact—
Impact cannot yet be scored: citation data is absent from the snapshot (h-index and i10-index both recorded as 0), so influence on subsequent literature — what this axis measures — remains unscored rather than estimated from the three 2026 preprints.
Focus0/100
Focus scores 0 because the field distribution is perfectly even: multimodal AI evaluation (scientific reasoning benchmarks), NLP (LLM evaluation for peer review), and machine learning (large language models) each account for exactly one third (0.333) of the record. Topic assignments — topic modeling (3), multimodal machine learning applications (2), plus scientometrics, NLP techniques, and multi-agent systems — confirm deliberate breadth across benchmarking, pretraining, and evaluation rather than a single narrow specialty.
Collaboration81/100
Collaboration scores 81 on strong network evidence: the OpenAlex snapshot records 58 distinct coauthors, and twelve collaborators recur across the catalog — including Tao Gui, Jiaqiang Li, Yajie Yang, and Zhiheng Xi — indicating sustained work within a larger research group. All three reviewed 2026 papers list Zhang as a middle author without corresponding status, consistent with a contributor role inside team-scale NLP projects.
Activity40/100
Activity scores 40: the snapshot records only 3 works in the last three years, all dated 2026, so the publication cadence is recent but thin. The counts-by-year data contains a single year (2026, with 3 outputs), leaving no visible multi-year track record; this more likely reflects an early or newly disambiguated OpenAlex profile than a measured slowdown.
Reproducibility85/100
Reproducibility is the strongest axis at 85, computed from 3 reviewed papers. NovGauge and NCP-ArchPreview carry high repro-readiness, while Sci-MMR is rated medium owing to LLM-judged coverage and unquantified variance. Consistent release of audited benchmarks and technical reports with explicit numbers (e.g., Stage-1 +2.45 versus Stage-2 +0.59) supports above-average replication readiness.
Research summary
Qi Zhang is a researcher at Fudan University whose visible output centers on rigorous evaluation infrastructure for large language models. The Evolve-reviewed catalog contains three 2026 arXiv preprints, each with Zhang in a middle author position: Sci-MMR, a 235-task argument-graph benchmark for multi-step, evidence-grounded scientific reasoning in multimodal agents; NovGauge, a fine-grained benchmark for diagnosing LLM capability in paper novelty assessment; and NCP-ArchPreview, a technical report on latent-space language models trained via next concept prediction at 8.9B parameters, claiming 1.95× token efficiency. This pattern matches the recorded topic mix of topic modeling (3), multimodal machine learning applications (2), scientometrics, NLP techniques, and multi-agent systems. OpenAlex signals show strong collaborative infrastructure: 58 distinct coauthors, a collaboration score of 81, and twelve recurring coauthors including Tao Gui and Zhiheng Xi. Reproducibility is the standout axis at 85, with two of the three papers rated high readiness and Sci-MMR rated medium. Output and impact remain unscored because citation data is unavailable in the snapshot; activity scores 40 on just three recent works, and the perfectly even one-third field split yields a focus score of 0.
Collaboration pattern
Zhang works in a large, stable collaborative network: 58 distinct coauthors appear in the OpenAlex record, and twelve collaborators recur — most prominently Tao Gui, Jiaqiang Li, Yajie Yang, and Zhiheng Xi. The collaboration score of 81 reflects this density. All three reviewed papers position Zhang as a middle author without corresponding status, suggesting a contributor within a team-scale NLP group at Fudan rather than a solo or lab-head signature pattern.
Research keywords
Highlights
Sci-MMR: a 235-task benchmark for evidence-grounded multimodal scientific reasoning
NovGauge: fine-grained diagnosis of LLM paper novelty assessment
NCP-ArchPreview: latent-space language models via next concept prediction
Timeline
- 2026Released Sci-MMR on arXiv, a 235-task argument-graph benchmark for multi-step evidence-grounded scientific reasoning in multimodal agents.
- 2026Released NovGauge on arXiv, a fine-grained benchmark diagnosing LLM capability in evidence-grounded paper novelty assessment.
- 2026Released the NCP-ArchPreview technical report, claiming 1.95× token-efficient pretraining at 8.9B parameters via next concept prediction.
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 drawn from OpenAlex only; no ORCID link is available, and 'Qi Zhang' is a common name, so this profile is name-resolved from public bibliographic data.
- Citation counts are unavailable (h-index and i10-index recorded as 0), so the output and impact axes are left unscored.
- The reviewed catalog covers three 2026 arXiv preprints only; peer-reviewed publications and earlier career output are not captured.
- All catalog papers list middle authorship with no corresponding-author role, so individual contributions within the papers cannot be attributed precisely.