Tsinghua researcher applying graph learning and NLP to scientific discovery and biomedical evidence synthesis
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
9
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-13) 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 snapshot reports an h-index of 9 and counts by year, but no verified per-year work list attributable to this specific researcher, so total publication volume and long-run output consistency cannot be measured reliably.
Impact—
Not yet scored: the data provides no per-paper citation counts, field-normalized percentiles, or venue-level quality signals, so this axis — meant to capture influence on downstream research — cannot be estimated from the current snapshot.
Focus0/100
The focus score is 0, reflecting a footprint split evenly between two field clusters — machine learning for scientific discovery and machine learning for biomedical evidence synthesis, each weighted 0.5 in the OpenAlex field distribution. Topic tags range from reinforcement learning in robotics (19 works) to NLP-based evidence synthesis, so the measured profile is broad rather than concentrated; part of this spread may reflect name conflation within the index rather than genuine breadth.
Collaboration36/100
Collaboration scores 36, based on a small, stable network: only 5 distinct coauthors across the 10 works recorded in the recent three-year window. The recurring partners are Fred Sun, Jingze Wang, and Minkun Xu. On both Evolve-reviewed 2026 papers Guo holds the last author position — conventionally a senior or supervisory role — yet is not marked corresponding author, and the narrow coauthor pool keeps the score modest.
Activity70/100
Activity scores 70, supported by sustained recent output: the OpenAlex counts-by-year series shows 5 works in 2019, 6 in 2021, and 3 in 2025, rebounding to 5 works in 2026, with 10 works inside the recent three-year window. The cadence is steady rather than explosive, with mild dips to 2 works in each of 2023 and 2024 followed by renewed momentum through 2025–2026.
Reproducibility20/100
Reproducibility scores 20 (reproMean 20 across 2 assessed papers). Both Evolve-reviewed 2026 papers carry a low reproducibility-readiness rating: one relies on distilled labels within a single domain, and the other validates on a single 140-query Cochrane benchmark. Reviewers nonetheless praised honest, well-controlled ablations, suggesting careful evaluation practice not yet matched by rerunnable artifacts.
Research summary
Shangqi Guo is a researcher affiliated with Tsinghua University (ORCID 0000-0003-3181-6881) whose cataloged work sits at the intersection of machine learning for scientific discovery (literature mining and graph learning) and machine learning for biomedical evidence synthesis (NLP/knowledge graphs) — OpenAlex weights these two clusters equally at 0.5 each. The Evolve-reviewed portfolio comprises two 2026 arXiv papers, both with Guo in the last author position: 'Time-Aligned Evolving Concept Graphs for Scientific Relation Forecasting', which aligns evolving semantics with temporal graph state at query time to improve relation forecasting and link prediction, and 'When More Evidence Hurts', which argues that deeper retrieval raises false positives on null biomedical causal queries and proposes publication-bias drift monitoring with KL-divergence-based principled stopping. Evolve judged both methodologically careful — honest, well-controlled ablations — while noting limits from distilled labels, a single 140-query Cochrane benchmark, and low reproducibility readiness (reproMean 20 across 2 assessed papers). The bibliometric snapshot shows an h-index of 9 (i10-index also 9), 10 works in the recent three-year window, 5 distinct coauthors, and steady output including 5 works recorded in 2026. Notably, the OpenAlex topic profile is dominated by robotics reinforcement learning (19 works), diverging from this NLP/graph portfolio — a mismatch that may reflect name conflation in the index.
Collaboration pattern
Guo's collaboration pattern is a small, stable core: 5 distinct coauthors across the 10 works in the recent three-year window, with Fred Sun, Jingze Wang, and Minkun Xu recurring most often. Both 2026 arXiv papers in the Evolve catalog place Guo in the last author position — conventionally the senior or supervisory seat — though neither marks corresponding authorship. The modest collaboration score of 36 mainly reflects team size rather than instability, since the same partners recur throughout the window.
Research keywords
Highlights
Deeper retrieval can hurt null-query causal search
Query-time alignment lifts scientific relation forecasting
Careful ablations, thin reproducibility
Across both Evolve-reviewed papers, reviewers praised honest, well-controlled ablations, yet both carry a low reproducibility-readiness rating (reproMean 20 across 2 papers). The pattern suggests theory-first designs that lean on distilled labels or single benchmarks rather than released, rerunnable pipelines.
Timeline
- 2026Published 'Time-Aligned Evolving Concept Graphs for Scientific Relation Forecasting' on arXiv as last author; Evolve's review found the query-time alignment of semantics and graph state effective, with well-controlled ablations but distilled labels and a single domain.
- 2026Published 'When More Evidence Hurts: Publication-Bias Drift and Principled Stopping for Biomedical Causal Search' on arXiv as last author; Evolve called the simple theory and honest ablations promising, on a single 140-query Cochrane benchmark.
- 2026OpenAlex snapshot records 5 works in 2026 and 10 works over the recent three-year window, with an h-index of 9 and i10-index of 9.
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.
- All bibliometrics come from an OpenAlex snapshot (identity: openalex); h-index 9 and i10-index 9 are quoted as provided, with no independent citation audit.
- The OpenAlex topic profile (top topic: Reinforcement Learning in Robotics, 19 works) and the 1996/1999 counts-by-year entries predate a plausible career window, suggesting possible name conflation; this profile is name-resolved from public bibliographic data plus the provided ORCID.
- Only two papers carry Evolve review verdicts; the quality of the broader portfolio is unassessed, and neither cataloged paper lists Guo as corresponding author.
- The output and impact dimensions are unscored because the underlying data is missing; no per-paper citation counts or verified work lists were available.