Harbin Institute of Technology ML researcher spanning PINN optimization, robust vision, and agentic agriculture.
3
papers here
0
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—
Output cannot be scored: the OpenAlex snapshot contains only three indexed works and no fuller publication history, so this axis — which would measure total production and its year-by-year trajectory — currently lacks sufficient data.
Impact—
Impact cannot be scored: the recorded h-index and i10-index are both zero with no citation counts by year, so this axis — which would measure citations and uptake of the work — awaits an accumulated citation record.
Focus—
Focus cannot be scored: five topics appear in the snapshot, each tied to a single work, so no concentration pattern is measurable yet; this axis would gauge how tightly the portfolio clusters around core research areas.
Collaboration60/100
Scoring 60, collaboration is the stronger axis in this snapshot: 19 distinct coauthors appear across the last three years, a wide network for only three indexed works. The cataloged arXiv papers place Jie Liu in middle and last author positions with no corresponding authorship, and recurring collaborators include Jing Xiao, Xinhai Chen and Qinglin Wang. The pattern suggests active, team-based participation across several small project groups rather than leadership of an independent group.
Activity40/100
Scoring 40, activity rests on a thin but current record: OpenAlex counts three works in the last three years, all assigned to 2026 in the counts-by-year data, with no indexed output shown for earlier years. This is consistent with a newly visible or early-career profile whose publishing rhythm cannot yet be judged over a longer horizon.
Reproducibility—
Reproducibility cannot be scored at profile level: the snapshot carries no aggregated artifact, code-release or benchmark metadata. This axis would measure whether reported results can be independently reproduced; Evolve's per-paper readiness notes are the only signal so far.
Research summary
Jie Liu is a machine learning researcher at Harbin Institute of Technology whose recent work, as captured in this OpenAlex-resolved profile, spans three fronts: optimization for scientific machine learning, robust computer vision, and agentic AI for smart agriculture. All three cataloged works are 2026 arXiv preprints, consistent with a record of three works in the last three years and an indexed h-index and i10-index of zero so far. The first paper identifies a gradient-update mismatch, showing that widely used optimizers such as Adam can break the conflict-freeness that gradient-surgery methods assume in physics-informed neural network training, and proposes a post-optimizer projection; Evolve judged the evidence well supported with high reproducibility readiness. A second paper, SAGE, tackles spurious correlations with group-label-free generative enhancement built on diffusion models and LoRA fine-tuning, reporting strong worst-group-accuracy gains. The third deploys FAIRY, a smart-agriculture agentic engine evaluated over 100 simulated full-season soybean operations, with the caveat that the study is digital-twin-only. With 19 distinct coauthors in three years and recurring collaborators such as Jing Xiao and Xinhai Chen, this looks like a broadly collaborative, early-stage researcher whose citation footprint has not yet accumulated; topic signals from remote sensing in agriculture to ethics of AI indicate unusually wide range for so few indexed works.
Collaboration pattern
Nineteen distinct coauthors in three years point to a broadly collaborative style. The three cataloged arXiv papers place Jie Liu in middle or last author positions without corresponding authorship, suggesting a contributing role within small teams. Recurring names include Jing Xiao, Xinhai Chen, Qinglin Wang, Menghan Jia and Zhiquan Lai, spanning optimization for scientific machine learning, robust vision and agentic agriculture — one small team per paper but a shifting, interdisciplinary cast across projects.
Research keywords
Highlights
Gradient-update mismatch: optimizers can break gradient surgery in PINN training
Group-label-free generative augmentation against spurious correlations
Agentic farm engine evaluated over 100 simulated soybean seasons
Timeline
- 2026Publishes the GUM analysis, showing that optimizers break gradient surgery's conflict-freeness in physics-informed neural network training and proposing post-optimizer projection.
- 2026Proposes SAGE, subpopulation-aware generative enhancement for mitigating spurious correlations without group labels.
- 2026Deploys and evaluates the FAIRY smart-agriculture agentic engine over 100 simulated full-season soybean farm operations.
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 only; no ORCID was available, and 'Jie Liu' is a common name, so this profile is name-resolved from public bibliographic data.
- Citation indicators are h-index 0 and i10-index 0 with a single counts-by-year entry (2026: 3), so impact cannot be meaningfully assessed.
- Only the three Evolve-reviewed 2026 arXiv preprints are in scope; other publications may exist outside this snapshot.
- Output, impact, focus and reproducibility scores are not computed due to insufficient underlying data.