SJTU researcher on LLM agents: verifiable computer-use environments and budget-aware RL search policies.
2
papers here
1
reviewed by Evolve
0
first-author
0
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—
The snapshot holds no publication counts, venue listings, or productivity totals, so output cannot be scored. Once indexed, this axis would measure total papers, publication venues, and cumulative volume over time.
Impact—
No citation counts, h-index, or citing works are available in the snapshot, so impact is unscored. This axis would measure how often and how widely the researcher's papers are cited across the literature.
Focus100/100
Focus scores a full 100 because the field distribution contains exactly one entry: machine learning / LLM agents. Both reviewed 2026 papers sit on this single line — RecreationWorld targets computer-use agent environments, while One Policy, Any Budget targets budget-aware search policies — yielding an unusually coherent agenda with no observed spread into adjacent topics in the reviewed record.
Collaboration73/100
Collaboration scores 73, backed by 38 distinct coauthors in the bibliometric record. Recurring names include Shuai Bai, Jiayong Deng, Sicheng Fan, Chang Gao, Mianqiu Huang, Yizhen Jiang, Yuheng Jing, Dehui Kong, Keliang Li, Ning Li, and Wanli Li. Both reviewed papers place Fu in a middle author slot, pointing to steady contribution inside sizeable, team-authored projects rather than solo-led work.
Activity—
Beyond the two reviewed 2026 preprints, the snapshot contains no dated publications, so year-by-year activity cannot be scored. This axis would track publishing cadence and recent output volume over consecutive years.
Reproducibility100/100
Reproducibility scores 100 as the mean across rated papers (reproMean 100), driven by RecreationWorld's high readiness — a five-platform recreation benchmark built to be scalable and verifiable. The base is thin, though: only one paper carries a rating (reproPapers 1), and the other reviewed paper is rated medium, so the perfect score reflects a single well-documented artifact rather than a track record.
Research summary
Yikun Fu is a machine-learning researcher affiliated with Shanghai Jiao Tong University whose reviewed work centers on large language model agents — systems that operate computers, call tools, and manage search budgets. Two 2026 arXiv preprints in the Evolve catalog, both with Fu in a middle author position, define that focus. RecreationWorld contributes a five-platform recreation benchmark with verifiable rewards plus an SFT pipeline for hybrid GUI/coding agents; Evolve's review calls the evaluation design rigorous while noting that transfer gains lack statistical controls. One Policy, Any Budget trains a single policy to internalize budget-aware search through scaffold curriculum reinforcement learning with GRPO; the review judges benchmarks and ablations strong but the gains modest and not yet thoroughly tested. The bibliometric radar mirrors this concentration: a focus score of 100 from a field distribution containing only machine learning / LLM agents, a collaboration score of 73 built on 38 distinct coauthors, and a reproducibility score of 100 — though that rests on a single rated paper. Output, citation impact, and year-by-year activity are not measurable in the current snapshot and remain unscored. Overall, this is an early-stage, tightly focused agent-systems contributor whose reviewed work emphasizes verifiable, reusable evaluation infrastructure.
Collaboration pattern
The record shows 38 distinct coauthors, and both reviewed papers place Fu in a middle author position — a pattern typical of team-driven agent research groups. Recurring collaborators include Shuai Bai, Jiayong Deng, Sicheng Fan, Chang Gao, Mianqiu Huang, Yizhen Jiang, Yuheng Jing, Dehui Kong, Keliang Li, Ning Li, and Wanli Li. No first or corresponding authorship appears in the reviewed catalog, so an independent leadership signal is not yet visible.
Research keywords
Highlights
RecreationWorld: five-platform verifiable environments for hybrid computer-use agents
One policy, any budget: internalizing budget-aware search with reinforcement learning
Timeline
- 2026Contributes as a middle author to RecreationWorld, an arXiv preprint presenting a five-platform recreation benchmark and SFT pipeline for hybrid GUI/coding agents.
- 2026Contributes as a middle author to One Policy, Any Budget, an arXiv preprint internalizing budget-aware search into a single policy via scaffold curriculum RL with GRPO.
- 2026Both reviewed papers fall under machine learning / LLM agents — the sole field in the record — anchoring the 100 focus score and a coherent agent-systems line.
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.
- Profile is name-resolved from public bibliographic data: no ORCID or country is recorded, and Yikun Fu is a common name, so identity matching carries some uncertainty.
- Output, impact, and activity are unscored because the snapshot contains no publication counts, citation metrics, or multi-year productivity data.
- The reproducibility score of 100 rests on a single rated paper (reproPapers = 1) and should not be read as a long-term track record.
- Evidence is limited to a bibliometric snapshot and two Evolve-reviewed arXiv preprints; work outside these sources is not represented.