Liyang Fan
Shenzhen Institutes of Advanced Technology · China
Early-career multimodal AI researcher at SIAT building agent evaluation benchmarks and efficient GUI grounding methods
2
papers here
2
reviewed by Evolve
2
first-author
0
corresponding
—
works · OpenAlex
2
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-20) 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: this snapshot lacks the per-year venue and publication-type breakdown the axis measures. The only available signals are 11 works in the last three years and an h-index of 2.
Impact—
No impact score yet: citation counts, citing works, and venue influence are absent from this snapshot. The axis would measure uptake of the two reviewed 2026 preprints and earlier records; only h-index 2 and i10-index 1 exist.
Focus0/100
The focus axis scores 0, and the raw data explains why: the field distribution splits evenly, 50/50, between machine learning (multimodal agents and evaluation) and computer vision (GUI agents and vision-language models). Recorded topics add topic modeling (4), NLP techniques (4), speech and dialogue systems (2), AI in law (2), and GAN-based image synthesis (1) — a broad, exploratory spread rather than a concentrated niche.
Collaboration56/100
Evolve scores collaboration at 56. The snapshot lists 16 distinct co-authors across 11 works from the last three years — about 1.5 per work — and ten recurring names, led by Yitai Li, Xinping Bi, Chi Wei, Guhong Chen, and Chenghao Sun. Both catalog papers are first-author but not corresponding, consistent with team-based projects where senior colleagues handle correspondence.
Activity72/100
Activity scores 72, the highest axis in this profile. OpenAlex records 11 works over the last three years, with by-year counts of 3 (2024), 5 (2025), and 3 (2026) — steady output for an early-stage researcher. Two of the 2026 works are first-author arXiv preprints already reviewed by Evolve, indicating the pace continues into benchmark- and methods-building.
Reproducibility55/100
Reproducibility scores 55, computed as the mean readiness across the 2 papers Evolve has reviewed (reproMean 55, reproPapers 2). Both ReFigBench and RankGround are rated medium: ReFigBench's benchmark-and-paired-study design aids replication but carries judge-family caveats, while RankGround reports strong, honestly analyzed results whose +5.5% headline Evolve flags for scrutiny.
Research summary
Liyang Fan is an early-career researcher at the Shenzhen Institutes of Advanced Technology (China) working across multimodal agents and vision-language systems. The snapshot, resolved from OpenAlex, records an h-index of 2 and i10-index of 1, with 11 works over the last three years (3 in 2024, 5 in 2025, 3 in 2026) and 16 distinct co-authors. Two 2026 first-author arXiv preprints reviewed by Evolve define the current profile. ReFigBench contributes a 1,000-figure benchmark for reconstructing scientific figures as editable PowerPoint artifacts, with a paired study showing that harness and workflow jointly decide agent quality. RankGround targets efficient high-resolution GUI grounding through a lightweight reranker guiding crop selection within a single vision-language-model call. Recorded topic counts — topic modeling (4) and NLP techniques (4) leading, with speech and dialogue systems (2), AI in law (2), and GAN-based image synthesis (1) — plus an even 50/50 field split indicate a deliberately broad, exploratory agenda rather than a settled niche. Evolve rates reproducibility readiness medium for both papers (mean 55 across 2 papers). The citation record is too thin for impact scoring, so this reads as an early-stage, output-oriented profile likely to shift as the preprints move through review.
Collaboration pattern
The OpenAlex snapshot lists 16 distinct co-authors across 11 works in the last three years, and Evolve's collaboration score is 56. Recurring names include Yitai Li, Xinping Bi, Chi Wei, Guhong Chen, Chenghao Sun, Haoxiang Yang, Qingwen Li, Kai Yan, Hong Li, and Bo Li. Both catalog papers are first-author but not corresponding, consistent with team-based projects where senior colleagues handle correspondence. The co-author list also contains 'Liyang Fan' itself, an identity-resolution artifact.
Research keywords
Highlights
ReFigBench: a 1,000-figure benchmark for editable PowerPoint reconstruction
Evidence that harness and workflow jointly decide agent quality
RankGround: single-call high-resolution GUI grounding via reranker crop selection
Timeline
- 2024OpenAlex's by-year counts record 3 works in 2024, the earliest year in the snapshot.
- 2025Publication count rises to 5 works in 2025, the highest annual total in the snapshot.
- 20263 works recorded in 2026, including two first-author arXiv preprints reviewed by Evolve: ReFigBench and RankGround.
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 come from a single OpenAlex snapshot; no ORCID is linked, and 'Liyang Fan' appears inside its own co-author list, so name resolution may merge or split records.
- Output and impact axes are unscored: citation counts beyond h-index 2 and i10-index 1 are not present in the data.
- Only two papers have been reviewed by Evolve; both are 2026 arXiv preprints and may not yet reflect peer-reviewed status.
- Career facts such as role, degree, and affiliation history are absent; topics and field distribution are inferred from bibliographic records only.