UCLA machine learning researcher working on efficient autonomous-driving inference and diffusion-model alignment
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
2
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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 not scored: the underlying data lacks a reliable cumulative publication count beyond the 14 recent-three-year works, so this axis — which would measure total paper volume and its spread across years and venues — is left unestimated.
Impact—
Impact is not scored: the current snapshot provides no per-paper citation data beyond h-index 2 and i10-index 1, which are too sparse to calibrate; this axis would measure citation-based influence such as citations per paper and per year.
Focus0/100
Focus scores 0 on the 0-100 scale: the record splits evenly between autonomous driving / efficient inference and diffusion-model alignment / RLHF (0.5 each in the field distribution), and spans five OpenAlex topic clusters — GANs and image synthesis (11), multimodal applications (8), media forensics (6), adversarial robustness (4), neural network applications (3). No single theme dominates yet, indicating an exploratory rather than concentrated agenda.
Collaboration51/100
Collaboration scores 51: 12 distinct coauthors across 14 recent-three-year works — roughly one new collaborator per paper — is mid-range breadth. Both reviewed papers are middle-author and non-corresponding, consistent with contributing inside larger teams. Recurring names include Xiaohuan Pei, Hengguang Zhou, Justin Cui, Yanchao Yang, Pichao Wang, and Cho-Jui Hsieh, suggesting at least one stable joint line of work.
Activity79/100
Activity scores 79, the researcher's strongest axis: 14 works in the last three years, and the by-year counts show a steep ramp — 1 (2022), 2 (2024), 1 (2025), then 11 in 2026. The 2026 cluster, which includes both Evolve-reviewed arXiv papers (SlackDrive and ReCAST), indicates a very high current submission pace for an early-career record.
Reproducibility38/100
Reproducibility scores 38 from Evolve's reviews of 2 papers (mean 38). SlackDrive rates low: the 21.7% headline latency figure appears in no table and key implementation details are missing. ReCAST rates medium: results are well executed but rely on a proprietary surrogate and report no significance tests. Released code and fuller artifacts would lift this axis.
Research summary
Yuanhao Ban is a machine-learning researcher at the University of California, Los Angeles whose OpenAlex-indexed record (identity resolved from public bibliographic data) shows an early-career, fast-accelerating profile: 14 works in the last three years, including 11 indexed in 2026 alone, against a cumulative h-index of 2 and i10-index of 1. Topic clusters concentrate on generative adversarial networks and image synthesis (11), multimodal machine learning applications (8), digital media forensic detection (6), adversarial robustness (4), and advanced neural network applications (3). The two papers Evolve has reviewed split evenly across the declared field distribution: autonomous driving / efficient deep learning inference (0.5) and machine learning for diffusion-model alignment / RLHF (0.5). SlackDrive (2026) proposes latency-feedback compute allocation for driving models — judged simple and well-analyzed, but its 21.7% headline appears in no table and reproducibility readiness is low. ReCAST (2026) assigns reward credit across diffusion timesteps via Rényi gains and Sinkhorn weighting — well executed, with modest gains, a proprietary surrogate, and no significance tests. Both are middle-author, non-corresponding arXiv preprints, consistent with a collaborative early-stage trajectory spanning 12 distinct recent coauthors and two rapidly growing subfields.
Collaboration pattern
Ban publishes within teams: 12 distinct coauthors across 14 recent-three-year works, and both catalog papers are middle-author arXiv preprints without corresponding-author status. The recurring roster includes Xiaohuan Pei, Hengguang Zhou, Justin Cui, Jiaqi Feng, Haoyu Xie, Tao Huang, Yihang Chen, and Kuei-Chun Kao, alongside Yanchao Yang, Pichao Wang, and Cho-Jui Hsieh. The pattern suggests steady group participation spanning two lines — efficient driving inference and diffusion-model alignment — rather than an independent agenda.
Research keywords
Highlights
SlackDrive: reusing runtime slack for adaptive driving inference
ReCAST: timestep-level reward credit assignment for diffusion RL
Timeline
- 2022First indexed work appears in the OpenAlex record (1 publication).
- 20242 works indexed — output remains intermittent.
- 20251 work indexed — a quiet year preceding the 2026 surge.
- 202611 works indexed in a single year, including both Evolve-reviewed arXiv papers (SlackDrive, ReCAST); both middle-author and non-corresponding.
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 OpenAlex only; the output and impact axes are unscored because citation data is insufficient in the snapshot.
- Only 2 papers carry Evolve reviews, both 2026 arXiv preprints; quality and reproducibility conclusions rest on this small sample with no venue information.
- The name is shown in pinyin; the Chinese character form is not confirmed by the source data.
- The frequent-coauthor list contained a duplicate entry (Cho-Jui Hsieh), deduplicated in this profile.