Machine-learning researcher at NUS working on efficient diffusion inference and vision-language interpretability
2
papers here
1
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
6
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—
No output score yet: the bibliometric snapshot does not include a verifiable complete works count for this profile. This axis would measure cumulative publication volume and its distribution across years and venues, so it is left unscored until fuller data is available.
Impact—
No impact score yet: citation totals and citation-quality signals for this profile are unavailable in the current snapshot. This axis would measure citation-based influence, including how often and by whom the work is cited; no estimate is attempted here.
Focus100/100
A maximal focus score of 100 reflects a tightly clustered topic profile: Domain Adaptation and Few-Shot Learning (count 10), Advanced Neural Network Applications (8) and Machine Learning and Data Classification (7) together account for 25 of 31 topic assignments, with smaller threads in topic modeling (3) and multimodal applications (3).
Collaboration50/100
A mid-range collaboration score of 50 reflects a compact network: 11 distinct co-authors appear across the record, and the frequent-coauthor list includes Cihan Acar, Tulika Mitra, Jun Luo, Zhongqi Yue and Weishi Wang. Both reviewed 2026 papers are multi-author efforts, though neither lists this researcher as corresponding author.
Activity67/100
An activity score of 67 rests on a steady recent cadence: OpenAlex counts 9 works in the last three years, with year-by-year counts of 3 (2022), 3 (2023), 2 (2024), 4 (2025) and 3 (2026), alongside an h-index of 6 and an i10-index of 4 — modest but continuous output with a 2025 peak.
Reproducibility55/100
Reproducibility scores 55, grounded in Evolve's review of 1 paper (reproMean 55, reproPapers 1); both catalogued 2026 arXiv papers carry a medium reproducibility readiness. The diffusion paper's unreported key thresholds and the VLM study's small sample are the main limiting factors flagged in review.
Research summary
Kuluhan Binici is a machine-learning researcher at the National University of Singapore whose work centres on making modern generative and vision-language models faster and more interpretable. The Evolve catalog carries two 2026 arXiv papers. The first, led as first author, introduces ADSC, a training-free adaptive step schedule controller for diffusion models that brings Stable Diffusion 1.5 down to roughly 19 inference steps while beating matched fixed-step baselines on CLIP-I and DINO; Evolve notes the evaluation is self-referential and key thresholds unreported. The second, a middle-author contribution, is a causal perturbation study showing that visual attention faithfulness in vision-language models is heterogeneous across three modes — novel, though a small sample and manual mode partition temper the conclusions. OpenAlex bibliometrics place the profile in a tightly focused machine-learning cluster: Domain Adaptation and Few-Shot Learning (10 records), Advanced Neural Network Applications (8) and Machine Learning and Data Classification (7) dominate, with smaller threads in topic modeling (3) and multimodal applications (3). Output is continuous — 9 works in the last three years, peaking at 4 in 2025 — with an h-index of 6 and an i10-index of 4, an early-career citation footprint. Collaboration remains compact (11 distinct co-authors), and reviewed reproducibility readiness is medium (mean 55).
Collaboration pattern
The visible collaboration footprint is compact: 11 distinct co-authors across recent work, with a recurring circle including Cihan Acar, Shivam Aggarwal, Siying Liu, Tulika Mitra, Xurui Song, Weishi Wang, Zhongqi Yue, Tao Bai, Hongxin Shao, Daniel Dahlmeier and Jun Luo. Both catalogued 2026 papers are collaborative and neither names this researcher as corresponding author, suggesting contributions within larger teams, consistent with a mid-range collaboration score of 50.
Research keywords
Highlights
Training-free diffusion inference cut to ~19 steps
Visual attention faithfulness in VLMs is heterogeneous
Concentrated early-career research focus
OpenAlex topics show a tightly clustered profile: Domain Adaptation and Few-Shot Learning (10), Advanced Neural Network Applications (8) and Machine Learning and Data Classification (7) account for 25 of 31 topic assignments, with continuous output of 9 works over the last three years.
Timeline
- 2022Earliest year in the available OpenAlex window: 3 works recorded, marking the start of the visible publication trajectory.
- 2023Publication cadence holds steady at 3 works, continuing the previous year's output level.
- 2024The lowest annual count in the available window, with 2 works recorded.
- 2025Peak productivity in the record: 4 works published in a single year.
- 2026Two arXiv papers enter the Evolve catalog: a first-author adaptive step schedule controller for efficient text-to-image diffusion, and a middle-author causal perturbation study of visual attention faithfulness in vision-language models.
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 indicators (h-index, i10-index, counts by year, topics, co-author counts) come from an OpenAlex-based snapshot; records are name-resolved from public bibliographic data and may merge or split same-name authors.
- Output and impact dimensions are unscored because the snapshot lacks complete works and citation data for this profile.
- The reproducibility score (55) is based on only 1 reviewed paper, and both catalogued entries are arXiv preprints rather than peer-reviewed venue publications.
- Field distribution is sparse (1 classified work: computer vision - efficient generative models), so field-level comparisons are limited.