Shanghai Jiao Tong University researcher working on LLM agents, on-device MoE inference, and agent skill orchestration
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
45
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—
Output scoring is not yet available: the record lacks the per-publication list needed to normalize annual volume. This axis would measure total and per-year paper counts across venues; only aggregate citation-window statistics (h-index 45) are currently on file.
Impact—
Impact scoring is not yet available: per-paper citation counts, citation distributions, and venue-level influence are missing. This axis would measure how strongly the work is cited relative to field norms, beyond the raw i10-index of 166.
Focus0/100
The focus score is 0, the most dispersed end of the scale. The field distribution splits evenly three ways — machine learning for LLM agents, mobile computer systems, and LLM agent orchestration at 0.333 each — while topic counts span auction theory (81), mobile ad hoc networks (68), privacy-preserving technologies (67), crowdsensing (53), and sensor networks (45). No single line of work dominates the record.
Collaboration76/100
The collaboration score is 76. The OpenAlex record lists 44 distinct coauthors, and twelve recur most often, including Yi Duan, Ying Liu, Zirui Tang, Haodong Chen, and Jun Zhou. All three Evolve-reviewed 2026 papers are multi-author works, with Fan Wu in middle or last author position and none as corresponding author, consistent with a contributor within larger teams rather than a solo investigator.
Activity100/100
The activity score is the maximum, 100. The record shows 83 works in the last three years, with countsByYear at 25 (2024), 27 (2023), 45 (2025), and 13 already in 2026; the historical peak was 40 works in 2014. Sustained output across two decades — from 2 works in 2001 to the current pace — marks an unusually consistent publishing trajectory.
Reproducibility32/100
The reproducibility score is 32, computed from three reviewed papers (reproPapers 3, reproMean 32). Evolve rates the recursive self-improvement survey and EStream as low readiness — ad hoc metrics and single-device testing respectively — and TROVE as medium, noting single-run evidence and unstated code release. The pattern suggests promising systems and agent work whose artifacts are not yet independently verifiable.
Research summary
Fan Wu is a researcher at Shanghai Jiao Tong University whose name-resolved OpenAlex record spans two decades of computer science publishing. The snapshot reports an h-index of 45, an i10-index of 166, 83 works over the last three years, and 44 distinct coauthors, with annual output peaking at 40 works in 2014 and reaching 45 in 2025. Topic counts concentrate on auction theory and applications (81), mobile ad hoc networks (68), privacy-preserving technologies (67), mobile crowdsensing (53), and energy-efficient wireless sensor networks (45). The three Evolve-reviewed papers, all 2026 arXiv releases, point toward large language models: a survey proposing a five-level autonomy taxonomy for recursive self-improvement, EStream on fast MoE prefill through expert virtualization on mobile NPUs, and TROVE on adaptive agent skill orchestration via trace-grounded route validation. Evolve rates their reproducibility readiness low, low, and medium respectively, yielding an aggregate reproducibility score of 32. Whether these agent- and systems-oriented papers mark a research pivot or reflect name-resolution overlap cannot be settled from the available data.
Collaboration pattern
The record lists 44 distinct coauthors, supporting a collaboration score of 76. Twelve collaborators recur most frequently — among them Yi Duan, Ying Liu, Zirui Tang, Haodong Chen, Jun Zhou, Yumou Liu, Bangrui Xu, Yukai Wu, Sidi Chen, Yuhan Zhou, Haoyu Wang, and Xiaoyou Yu. In the three reviewed 2026 papers Fan Wu appears as a middle or last author and never as corresponding author, suggesting work conducted within larger project teams.
Research keywords
Highlights
A five-level autonomy taxonomy for recursive self-improvement
First full-NPU MoE prefill on a smartphone
Runtime route editing for agent skill orchestration
Timeline
- 2026Posts a survey proposing a five-level autonomy taxonomy for recursive self-improvement in LLM-based agents.
- 2026Releases EStream, achieving the first full-NPU MoE prefill on a smartphone through shared expert graphs and flash streaming.
- 2026Introduces TROVE, an adaptive agent skill orchestration method built on trace-grounded route validation and editing.
- 2026All three Evolve-cataloged works appear as arXiv preprints in the same year, spanning agent surveys, mobile systems, and orchestration.
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 an OpenAlex name-resolved record; 'Fan Wu' is a common name, so the long-running auction/crowdsensing topics and the 2026 LLM-era papers may partly reflect identity mixing.
- Only aggregate citation statistics (h-index 45, i10-index 166) are on file; per-paper citations and full publication lists are absent, so the output and impact axes are unscored.
- All three reviewed papers are 2026 arXiv preprints and have not undergone peer review.
- The reproducibility score (32) is computed from only three reviewed papers and may not represent the wider portfolio.