University of Hong Kong researcher in scheduling optimization and multimodal reasoning benchmarks
3
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
16
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 is unscored: the snapshot lacks the aggregate publication data this axis requires. It would measure cumulative publication volume and productivity trajectory; only per-year counts (from 2 works in 2017 to 6 in 2026) are visible here.
Impact—
Impact is unscored: citation-based aggregates are not in the snapshot. The axis would measure citation counts, field-normalized influence and scholarly uptake; only h-index 16 and i10-index 20 are recorded as partial indicators.
Focus100/100
Focus scores 100. The machine-classified field distribution for the most recent works lists a single field — artificial intelligence (multimodal model evaluation) — so concentration metrics read as maximal. Catalogued output supports coherent clustering around scheduling optimization and benchmark design, though OpenAlex topic counts also surface advanced thermoelectric materials (10) and cardiac and mental health (8), a spread that may partly reflect name-merging in the underlying record.
Collaboration58/100
Collaboration scores 58, a mid-range value. The last three years produced 18 works with 17 distinct coauthors — roughly one new collaborator per work — and all three 2026 catalogued papers place Liu in the middle of the author list without corresponding-author roles. Recurring names include Shaochong Lin, Max Z. J. Shen, Jianfeng Gao and Ruohan Zhang, indicating small stable clusters rather than large-lead teams.
Activity86/100
Activity scores 86, among the higher readings in this profile. The provided countsByYear series shows 18 works in the last three years (5 in 2024, 7 in 2025, 6 in 2026), a clear step up from the 2-work years of 2017 and 2019, with the decade peak at 9 works in 2021. The current pace is near-peak and sustained.
Reproducibility100/100
Reproducibility scores 100, drawn from a reproducibility mean of 100 across 1 reviewed paper. That paper, MindTopo, carries a high reproducibility-readiness rating: Evolve credits its strong procedural design — a rigorous 13-task benchmark — while noting no significance tests were shown. With only one paper scored, this perfect reading should be read as promising but early.
Research summary
Anbang Liu is a researcher at the University of Hong Kong whose bibliometric record spans optimization and scheduling algorithms, multimodal and embodied AI evaluation, and healthcare operations. The 2026 arXiv record includes two learning-based scheduling contributions for semiconductor manufacturing — congestion-aware route scheduling for fab material control systems and trajectory-initialized neural double Q-routing for large-scale overhead hoist transport — alongside MindTopo, a 13-task benchmark asking whether foundation models can reason in topological space. Evolve's verdict: multimodal large language models reason about topology far better than they plan; the procedural design is strong, but no significance tests are shown, and reproducibility readiness is rated high. The underlying OpenAlex data records h-index 16 and i10-index 20, with 18 works in the last three years against a ten-year series that peaked at nine works in 2021, and 17 distinct coauthors. Topic counts center on scheduling and optimization algorithms (10) and advanced thermoelectric materials and devices (10) — a breadth that may reflect name-merging in public bibliographic data. Output and impact dimensions lack sufficient data for scoring; caveats apply.
Collaboration pattern
Liu's recent record is collaborative but not leadership-heavy: 18 works in the last three years with 17 distinct coauthors, and all three 2026 catalogued papers list Liu in a middle-author position without corresponding status. The collaboration dimension scores 58, and recurring coauthors include Shaochong Lin, Max Z. J. Shen, Hao Yin, Qineng Wang, Jianwen Lyu, Jianfeng Gao and Ruohan Zhang — a mix spanning operations, optimization and AI contexts.
Research keywords
Highlights
MindTopo: a 13-task test of whether foundation models reason topologically
Learning-assisted congestion-aware routing for semiconductor fabs
A 2026 arXiv paper on route scheduling for fab material control systems, applying learning-assisted, congestion-aware scheduling to semiconductor manufacturing logistics; listed in the catalog without an Evolve review yet.
Trajectory-initialized neural double Q-routing for overhead hoist transport
A 2026 arXiv paper proposing trajectory-initialized neural double Q-routing for large-scale overhead hoist transport systems, extending reinforcement-learning routing to fab-scale material handling; not yet reviewed by Evolve.
Timeline
- 2017Earliest year in the provided countsByYear window, with 2 works recorded.
- 2021Peak of the ten-year window: 9 works, the highest annual count on record.
- 2023Annual output dipped to 3 works before recovering in subsequent years.
- 2026Six works so far, including all three catalogued arXiv papers: MindTopo, congestion-aware fab route scheduling, and neural double Q-routing for OHT systems.
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 snapshot is OpenAlex-sourced (identity: openalex); citation aggregates beyond h-index 16 and i10-index 20 were not provided, so the output and impact axes remain unscored.
- OpenAlex topics span thermoelectric materials, scheduling and cardiac/mental health, which may reflect name-merging of homonymous authors; this profile is name-resolved from public bibliographic data.
- Only MindTopo has Evolve review data; the other two 2026 arXiv papers carry no verdict, keywords or reproducibility rating.
- The reproducibility score of 100 rests on a single reviewed paper (reproPapers: 1) — a very small base.