LLM systems researcher at City University of Hong Kong: agentic RL for software engineering, efficient MoE inference.
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
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—
This axis would measure annual publication volume and growth trend, but the OpenAlex snapshot lacks a complete works-by-year series — only a 2026 figure of 37 works and a three-year count are present — so no output score can be computed yet.
Impact—
Impact scoring requires citation counts, h-index trajectory, and per-paper influence; the snapshot records h-index 0 and i10-index 0 with no citation distribution, so this axis is left unscored rather than estimated.
Focus100/100
Focus scores 100: the snapshot's topic distribution is tightly concentrated, led by generative adversarial networks and image synthesis (14 papers) and multimodal machine learning applications (13), with three topics tied at 11 — topic modeling, advanced neural network applications, and parallel computing and optimization techniques — all clustered in applied deep learning and ML systems, matching the two reviewed 2026 papers on LLM agents and efficient MoE inference.
Collaboration58/100
Collaboration scores 58. The snapshot lists 17 distinct coauthors in the recent window and 12 recurring collaborators, including Jie Zhao and Ziyu Jiang. On the two reviewed papers Lin Qu holds middle and last author positions without corresponding authorship, consistent with a contributing or supporting-senior role in larger teams rather than solo or lead-driven work.
Activity100/100
Activity scores 100, the maximum. The snapshot records 37 works in the last three years, and the by-year counts attribute all 37 to 2026 alone, indicating an exceptionally dense recent output burst. Both Evolve-reviewed papers are dated 2026, reinforcing that current research activity — on software engineering agents and MoE inference — is concentrated in this single, very recent year.
Reproducibility20/100
Reproducibility scores 20, derived from Evolve's reproducibility assessment (reproMean 20 across one assessed paper). The software engineering agents paper is rated low readiness, with a missing compute-matched baseline flagged; the PCoMoE inference paper is rated medium. The sample is small, so this score should be read as an early, low-confidence signal.
Research summary
Lin Qu is a researcher at City University of Hong Kong whose profile, name-resolved from OpenAlex bibliographic data, sits at the intersection of applied deep learning and large language model systems. The snapshot records 37 works in the last three years, all attributed to 2026 in the by-year counts, alongside a citation record of h-index 0 and i10-index 0 — a very early bibliometric footprint. Topic frequencies concentrate on generative adversarial networks and image synthesis (14), multimodal machine learning applications (13), topic modeling (11), advanced neural network applications (11), and parallel computing and optimization techniques (11), with 17 distinct coauthors. Two Evolve-reviewed 2026 arXiv papers anchor the catalog. On the software-engineering-agents paper 'One to More, More to One,' Lin Qu is the last author; Evolve reads the three category RL experts merged via multi-teacher distillation as promising but notes a missing compute-matched single-policy baseline. On 'PCoMoE,' a mixture-of-experts inference paper, Lin Qu is a middle author; Evolve finds the engineering coherent while judging that the abstract's '10% accuracy' phrasing overstates the +1.6–5.8-point gains in Table 1. Radar scores: activity 100, focus 100, collaboration 58, reproducibility 20; output and impact remain unscored.
Collaboration pattern
The snapshot records 17 distinct coauthors in the recent three-year window and 12 recurring collaborators, including Jie Zhao, Ziyu Jiang, Suhang Zheng, Minghui Shan, Xiaoxiao Xu, Ziyan Gan, Fangxin Liu, Chenyang Guan, Junjie Wang, Ning Yang, Haomin Li, and Xiang Li. On both reviewed 2026 papers Lin Qu appears in non-corresponding middle and last positions, a contributing pattern in larger teams; Evolve's verdict ties the SWE-agents work to Alibaba.
Research keywords
Highlights
Last-author role on category-aware expert training for SWE agents
Middle-author role on PCoMoE MoE inference acceleration
Timeline
- 2026Co-authored 'One to More, More to One: Category-Aware Iterative Expert Training for Software Engineering Agents' (arXiv) as last author, a paper on agentic reinforcement learning for repository-level software engineering.
- 2026Co-authored 'PCoMoE' (arXiv) as a middle author, proposing fine-grained expert path composition for more efficient mixture-of-experts LLM inference.
- 2026Per the OpenAlex snapshot, all 37 of the last-three-years works are attributed to this year, with 17 distinct coauthors recorded in the window.
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 a single OpenAlex snapshot; no other citation databases were cross-checked.
- h-index and i10-index are both recorded as 0, and the by-year work counts cover only 2026 (37 works), so the output and impact axes are left unscored.
- 'Lin Qu' is a name resolved from public bibliographic data; identity disambiguation (including possible record merging or splitting) has not been independently verified.
- The reproducibility score of 20 rests on a single Evolve-assessed paper and should be treated as low-confidence.