Kuaishou researcher bridging recommender systems and LLM agents: user simulation, prompt optimization, bandit methods
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
23
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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: complete per-work output records are unavailable for this identifier, so publication volume cannot be computed reliably. This axis would measure total works, yearly output distribution, and authorship roles once full bibliometric coverage is available.
Impact—
Not yet scored: citation linkage for this identifier is incomplete, so influence cannot be computed. This axis would capture citation counts, citing works, and per-paper impact; the h-index of 23 and i10-index of 35 alone are insufficient to score it.
Focus0/100
The focus score is 0, reflecting an even 0.5/0.5 split between LLM agent systems and NLP dialogue/user simulation in the current field distribution. The OpenAlex topic profile is itself broad: recommender systems and search behavior (69 works each), bandit algorithms (35), topic modeling (21), and neural network applications (9). Rather than one narrow line, the recent 43 works span two distinct research programs.
Collaboration62/100
Collaboration scores 62, based on 22 distinct coauthors in the recent window and a named frequent-coauthor pool of 12, including Zibo Zhao, Zhongyuan Wang, Ruiming Tang, and Kun Gai. On the two reviewed papers he holds no first-author seat — last author on TRACER and a middle author on RobustSGPO, neither corresponding — consistent with a senior-industry contributor working inside large Kuaishou teams rather than leading small groups.
Activity100/100
Activity is the standout axis at 100. The recent three-year window contains 43 works, and countsByYear peaks sharply in 2026 with 36 records — more than any prior year and nearly half the recent window — after single digits in 2023-2025 (3, 3, 4) and a 2022 low of 1. Both reviewed 2026 papers sit inside this surge, indicating an intense current publishing phase.
Reproducibility38/100
Reproducibility scores 38, the mean readiness across 2 papers assessed by Evolve. TRACER is rated medium: results are strong, but benchmark claims rest on thin statistics. RobustSGPO is rated low: gains are consistent, yet evidence comes from a single workflow with no significance testing. Neither reviewed paper is corresponding-authored by him, so reporting practices likely reflect team norms.
Research summary
Wenwu Ou is a researcher at Kuaishou (China) whose OpenAlex-indexed work spans recommender systems, information retrieval, and, more recently, LLM-based agents and dialogue. The bibliometric snapshot records an h-index of 23 and an i10-index of 35, with topic mass concentrated in recommender systems and search behavior (69 works each), bandit algorithms (35), topic modeling (21), and neural network applications (9). Output is heavily front-loaded into 2026: 36 works recorded that year against 43 across the trailing three years combined, driving an activity score of 100. His current field distribution splits evenly (0.5/0.5) between LLM agent systems and NLP dialogue/user simulation. Two 2026 arXiv papers are reviewed in the Evolve catalog: TRACER, aligning a 7B multi-turn user simulator to real user decision trajectories via DTW-rewarded reinforcement learning (last author), and RobustSGPO, adding search-space control to prompt optimization for agent harness evolution (middle author). Evolve's verdicts credit strong empirical gains while flagging thin benchmark statistics and absent significance testing, yielding a reproducibility score of 38 across 2 assessed papers. Collaboration is active — 22 distinct recent coauthors including Zibo Zhao, Ruiming Tang, and Kun Gai — scoring 62. Output and impact axes remain unscored pending fuller bibliometric coverage.
Collaboration pattern
His collaboration footprint is industry-scale and team-based: 22 distinct coauthors in the recent window and a recurring pool of 12 — Zibo Zhao, Jijun Shi, Mo Zhou, Zhongyuan Wang, Shifu Bie, Yunfei Zhang, Xuanting Zhou, Xiangyu Wu, Bin Liu, Ruiming Tang, Kun Gai, and Geng Chen — consistent with Kuaishou project groups. On the two reviewed 2026 papers he takes neither first nor corresponding authorship (last and middle positions), fitting a senior-contributor role inside larger teams.
Research keywords
Highlights
TRACER: behavior-consistent multi-turn user simulators via DTW-rewarded RL
RobustSGPO: search-space control for agent harness evolution
Timeline
- 2018Recorded output rises from 2 works in 2017 to 10 in 2018, the first visible ramp in the countsByYear series.
- 2019An early peak of 16 works — the highest single-year count before 2026 — spanning recommender systems, search behavior, and bandit algorithm topics.
- 2022Output falls to a single work, the lowest point of the recorded window, followed by single-digit years (3, 3, 4) through 2023-2025.
- 2026A surge to 36 recorded works — nearly half the 43-work recent three-year window — including both Evolve-reviewed arXiv papers on user simulation and agent-harness prompt optimization.
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.
- Bibliometric figures (h-index, yearly counts, topics, coauthors) come from an OpenAlex-based snapshot; the output and impact dimensions are unscored because the underlying data is insufficient.
- Only 2 papers from the 2026 arXiv batch carry Evolve reviews; earlier works are counted statistically but not individually assessed.
- The 2026 count of 36 likely reflects partial-year and preprint indexing and may be revised as coverage settles.
- Profile is name-resolved from public bibliographic data (OpenAlex, ORCID 0009-0004-2437-6835); affiliation is Kuaishou (China) and no country field is separately recorded.