Yanlin Wang

Sun Yat-sen University · China

Unclaimed · AI-structured profileORCID OpenAlex

Researcher at Sun Yat-sen University working on LLM-agent evaluation, SWE benchmarks, and repo-level code generation

papers here

reviewed by Evolve

first-author

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—

Output is unscored: the current data provides no reliable publication-volume measure for this profile. This axis would summarize total works, per-year output, and venue spread; until such data is available, no estimate is offered.

Impact—

Impact is unscored: citation evidence is insufficient in the snapshot (OpenAlex reports h-index 1 and i10-index 0). This axis would measure citation uptake and scholarly influence of the work; no estimate is made.

Focus100/100

Focus scores 100/100: the field distribution assigns this profile to a single field, software engineering (AI for code), and topic counts cluster tightly within it — Software Engineering Research (25), Software Testing and Debugging Techniques (15), Topic Modeling (13), Software Engineering Techniques and Practices (11), and Information Retrieval and Search Behavior (5). All three Evolve-reviewed papers sit on the same LLM-agent evaluation and repository-level code generation line.

Collaboration55/100

Collaboration scores 55/100. The snapshot records 24 works with 15 distinct coauthors over the last three years, and the recurring-coauthor list includes Mingwei Liu, Kefeng Duan, Dewu Zheng, Jiachi Chen, Ensheng Shi, Xilin Liu, Yuchi Ma, Zibin Zheng, Terry Yue Zhuo, Jianxing Yu, and Xin He. All three cataloged papers place Wang in a middle author position without corresponding status, consistent with team-based work rather than solo leadership.

Activity94/100

Activity scores 94/100. The counts-by-year series runs 1 work in 2022, 3 in 2025, and 21 in 2026, and the snapshot credits 24 works over the last three years — a steep ramp concentrated in 2026, when the three Evolve-reviewed arXiv preprints appeared. This is an early-stage but rapidly accelerating publishing record.

Reproducibility55/100

Reproducibility scores 55/100, computed from a single rated paper (reproMean 55, reproPapers 1). All three Evolve-reviewed papers carry a medium repro-readiness tag, and the verdicts flag methodological cautions — single-run statistics in SWE-Gate, small agent pools in the IRT evaluation, and missing significance tests or selective-RAG baselines in the retrieval work.

Research summary

Yanlin Wang is a software engineering researcher at Sun Yat-sen University whose OpenAlex-linked profile concentrates almost entirely on software engineering with AI for code: the field distribution lists that single field, and topic counts are led by Software Engineering Research (25), Software Testing and Debugging Techniques (15), Topic Modeling (13), Software Engineering Techniques and Practices (11), and Information Retrieval and Search Behavior (5). Output is recent and steep: the counts-by-year series records 1 work in 2022, 3 in 2025, and 21 in 2026, with 24 works over the last three years and 15 distinct coauthors. Three 2026 arXiv preprints reviewed by Evolve define the current agenda. SWE-Gate is a 303-instance benchmark arguing that passing functional tests is insufficient for software engineering agents. A trajectory-aware evaluation method (PTA-IRT) inserts trajectory summaries as privileged information into item response theory for cheaper agent benchmarking, winning consistently on 4 benchmarks. Critical token-aware retrieval for repository-level code generation reports 8.4–15.4% relative Pass@5 gains. All three papers list Wang as a middle author rather than corresponding author, consistent with contributing inside an active team. Citation-based impact and output scores are not computed for this profile; see caveats.

Collaboration pattern

Coauthorship is team-based: 15 distinct coauthors across 24 works in the last three years (collaboration score 55). The recurring list centers on a stable core group — Mingwei Liu, Kefeng Duan, Dewu Zheng, Jiachi Chen, Ensheng Shi, Xilin Liu, Yuchi Ma, Zibin Zheng, Jianxing Yu, Xin He — plus external collaborator Terry Yue Zhuo. On all three Evolve-reviewed papers Wang appears as a middle author rather than corresponding author, suggesting contribution within a led project rather than sole direction.

Research keywords

LLM agentsSWE-Benchbenchmark constructionrepository-level code generationretrieval-augmented generationitem response theoryefficient evaluationcode reviewsoftware testingtrajectory summarization

Highlights

SWE-Gate: functional test passage is not enough for SE agents
A 303-instance benchmark arguing that software engineering agents must satisfy review-style constraints beyond passing functional tests. Evolve finds the evidence supportive but flags single-run statistics and the realism of synthetic tasks as points needing scrutiny.
Cheaper SWE-agent evaluation with trajectory-aware IRT
PTA-IRT uses trajectory summaries as privileged information inside item response theory to cut the cost of benchmarking SWE agents, with consistent wins on 4 benchmarks; Evolve notes that small agent pools limit certainty.
Critical token-aware retrieval for repository-level code generation
A token-triggered, on-demand retrieval mechanism for repository-level code generation, reporting 8.4–15.4% relative Pass@5 gains; Evolve deems the gains plausible but notes the absence of significance tests and selective-RAG baselines.

Timeline

2022Earliest year in the OpenAlex counts-by-year series, with 1 recorded work — the starting point of this profile's visible publication record.
20253 works recorded in the counts-by-year series, a modest step up as output began to build.
202621 works recorded, the peak of the series; Evolve reviews three arXiv preprints from this year — SWE-Gate, trajectory-aware agent evaluation (PTA-IRT), and critical token-aware retrieval for repository-level code generation.

Papers in the Evolve catalog

Frequent co-authors

Mingwei Liu×3 Kefeng Duan×2 Dewu Zheng×2 Jiachi Chen×2 Ensheng Shi×2 Xilin Liu×2 Yuchi Ma×2 Zibin Zheng×2 Terry Yue Zhuo×1 Jianxing Yu×1 Xin He×1 Jiachi Chen×1

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.

Bibliometric indicators come from an OpenAlex snapshot only; the reported h-index (1) and i10-index (0) are quoted verbatim from that source, and no citation-based impact score is computed.
"Yanlin Wang" is a common name; this profile is name-resolved from public bibliographic data (OpenAlex identity linked to ORCID 0000-0001-7761-7269) and may merge or split records.
Output and impact radar axes are unscored (dimMissing): the underlying data is insufficient and no estimate is substituted.
The reproducibility score (55) rests on a single rated paper (reproPapers 1), so it is a thin-base indicator.