Bowen Zhou

Output—

Not scored: this OpenAlex record lists h-index 0 and i10-index 0 with works captured only for 2026, so indexed output is too incomplete to support a reliable productivity score; linking an ORCID and back-filling the publication record would be needed first.

Impact—

Not scored: citation counts, citing works, and venue-level influence are absent from the snapshot (the record reports h-index 0), and all cataloged papers are 2026 preprints, so research impact cannot yet be assessed on available data.

Focus0/100

Scores 0 because the work is spread rather than concentrated: fieldDist splits evenly at roughly 33% each across AI for science/computational chemistry, large language models, and reaction data infrastructure, and topics scatter over at least five clusters (computational drug discovery 5, ML in materials science 5, planning 2, multimodal 2, topic modeling 1). Some of this may reflect name-resolution noise in the OpenAlex record.

Collaboration86/100

Scores 86: OpenAlex records 72 distinct coauthors across the 7 recent works, indicating large-team, lab-scale collaboration rather than solo research. A recurring core includes Biqing Qi, Dahua Lin, Dong Li, Tong Zhu, Kaiyan Zhang, and Junqi Gao. On the three cataloged papers Zhou holds middle or last author positions and is never the corresponding author, consistent with specialist roles inside bigger Shanghai AI Laboratory groups.

Activity61/100

Scores 61: recent3yWorks is 7 and countsByYear contains only 2026 with 7 publications, meaning every captured output landed in a single year with no earlier record indexed. The pattern reads as an active 2026 burst across three fields rather than sustained low output, and the moderate score largely reflects how thin the multi-year record is, not inactivity.

Reproducibility58/100

Scores 58, the reproMean over 3 repro-scored papers, and the spread is wide. The NCP-ArchPreview report is rated high reproducibility readiness; the agentic mechanism-discovery paper is medium, with n=3 case studies and LLM-judged evaluation; DianShi-RxnDB is low, with 24M-reaction pipeline details deferred and no downstream benchmark. The pattern suggests willingness to release artifacts, with rigor concentrated in the pretraining work.