Early-career Fudan University researcher in automated LLM fine-tuning, LLM agents, and human-AI collaboration
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
2
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—
This axis measures publication volume, cadence, and authorship role. It is not scored yet: the record is thin, and both cataloged 2026 papers place this researcher in a middle author position, so a stable output score needs fuller data.
Impact—
This axis would measure citation-based influence — total citations, h-index, and adoption of methods by others. The snapshot is too early to score: h-index is 2 and i10-index is 0, so impact is left unscored rather than estimated.
Focus0/100
The focus score is 0 because effort is split evenly: the field distribution is 50% machine learning (automated LLM fine-tuning, LLM agents) and 50% AI agents and human-AI collaboration. Topic data reinforces this breadth — topic modeling (4 papers), AI-based problem solving and planning (3), domain adaptation and few-shot learning (3), plus speech and dialogue systems (2) and educational technology and pedagogy (2).
Collaboration85/100
Collaboration scores 85, grounded in 70 distinct coauthors in the OpenAlex record — a wide network for an early-stage profile. Recurring partners include Haoran Zhao, Wei Du, Dingwen Yang, Jixuan Huang, Lingyong Fang, Ya Guo, Tao Gui, Qi Zhang, and Xuanjing Huang, a dense repeat cluster around the stated Fudan affiliation; both 2026 catalog papers continue this multi-author, team-based working mode.
Activity61/100
Activity scores 61. The counts-by-year record shows 2 works in 2021, 1 in 2022, 2 in 2025, and 5 in 2026 — the largest year so far and a steep recent ramp, with 7 works in the last three years. The 2023-2024 gap is not explained in the provided data.
Reproducibility38/100
Reproducibility scores 38, the mean across 2 Evolve-assessed papers. The SAGE fine-tuning paper is medium readiness: it reports Avg@3 gains from 3.2% to 15.6% over 9 tasks, but from only 3 runs with no significance tests. Atria Dawn is low readiness, with cherry-picking risk flagged in its benchmark table and self-reported attribution numbers.
Research summary
Junlin Shang is an early-career researcher at Fudan University whose recorded work splits evenly between two streams: machine learning centered on automated LLM fine-tuning and LLM agents, and AI agents with human-AI collaboration — each 50% of the field distribution in the OpenAlex snapshot. Two 2026 arXiv papers appear in the Evolve catalog, both in a middle author position. SAGE introduces strategy accumulation and guided execution for automated fine-tuning, maintaining an experience repository so that Monte Carlo Tree Search and a multi-agent pipeline can reuse prior strategies; it lifts Avg@3 performance across 9 tasks from 3.2% to 15.6%, though on 3 runs with no significance tests. Atria Dawn is a model release report toward agentic superintelligence whose human-AI collaboration study is its most substantive part, while its benchmark table and self-reported attribution numbers carry cherry-picking risk. Beyond these, topic data shows strengths in topic modeling, AI-based problem solving and planning, and domain adaptation and few-shot learning, with further work in speech and dialogue systems and educational technology. The bibliometric profile records 70 distinct coauthors (collaboration score 85), 7 works in the last three years with 5 recorded for 2026, and thin citation data (h-index 2, i10-index 0), consistent with an accelerating early-career trajectory.
Collaboration pattern
The record lists 70 distinct coauthors, and collaboration scores 85. The most frequent partners — Haoran Zhao, Wei Du, Dingwen Yang, Jixuan Huang, Lingyong Fang, Ya Guo, Tao Gui, Qi Zhang, Xuanjing Huang, Honglin Guo, and Yicheng Chen — form a dense recurring cluster around the Fudan affiliation. Both 2026 catalog papers use large-team authorship with this researcher in a middle position and not corresponding author, suggesting a contributor role inside a lab-style group rather than an independent lead.
Research keywords
Highlights
SAGE makes automated LLM fine-tuning cumulative
Atria Dawn: agentic superintelligence report with a human-AI collaboration core
Mixed reproducibility profile at an early stage
Across the 2 Evolve-assessed papers the reproducibility mean is 38: SAGE reaches medium readiness with its reported numbers, while Atria Dawn is rated low, signaling heavier dependence on self-reported benchmarks.
Timeline
- 2021First recorded works appear in the OpenAlex record (2 works).
- 2022Output slows to a single recorded work.
- 2025Publishing resumes with 2 recorded works, opening the current phase.
- 2026Most active year: 5 recorded works, including two Evolve-cataloged arXiv papers — SAGE on automated LLM fine-tuning and Atria Dawn on agentic superintelligence.
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 snapshot; no ORCID is linked, so name-based identity resolution should be treated with care.
- Output and impact are unscored: citation data is thin (h-index 2, i10-index 0) and the 2023-2024 gap in counts-by-year is unexplained in the provided data.
- The reproducibility mean (38) rests on only 2 Evolve-reviewed papers.
- Only 2 papers by this researcher are reviewed in the Evolve catalog; the profile may not reflect the full publication record.