Yifan Yang

Shanghai Jiao Tong University · Germany

Unclaimed · AI-structured profileOpenAlex

ML researcher at Shanghai Jiao Tong University: NAS performance prediction and multimodal intent understanding

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

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 snapshot lacks reliable per-venue publication records for this identity. This axis would measure total works, venue quality, and first/corresponding-author share once publication records are disambiguated.

Impact—

Impact is not scored: citation-linked signals (h-index, i10-index, citation counts) are absent for this identity in the snapshot. This axis would measure citations, uptake, and the influence of the work as it accumulates over time.

Focus8/100

Focus scores 8/100, the lowest axis on the radar, reflecting a wide topical footprint: OpenAlex topics span five clusters, led by Multimodal Machine Learning Applications (16 works) and GAN-based image synthesis (8), while Evolve's field split assigns 66.7% to neural architecture search and 33.3% to speech and audio-visual multimodal learning. The two reviewed threads are related but distinct, and part of this dispersion may stem from name-resolution noise.

Collaboration65/100

Collaboration scores 65/100. The snapshot lists 25 distinct co-authors and a recurring pool of 12 frequent collaborators, including Zhaoyan Wang, Zheng Gao, Xiaoyu Li, Jiaojiao Jiang, Qi Chen, Yunfei Chu, Haolin He, Zihan Liu, Yuxuan Wang, Ziyang Ma, and Ruiyang Xu. Authored contributions sit in team settings: first author on both NAS papers but middle author on the ODU-Bench benchmark, indicating method-lead roles inside larger group efforts rather than solitary work.

Activity100/100

Activity scores 100/100, the strongest axis: the snapshot attributes 30 works to this identity in 2026, with recent3yWorks also at 30 — an unusually high single-year volume. All three Evolve-reviewed papers are 2026 arXiv preprints, two as first author. This figure may partly reflect name-resolution breadth, so it is best read as identity-level record volume rather than verified solo output.

Reproducibility70/100

Reproducibility scores 70/100, a mean computed across the three Evolve-reviewed papers. ODU-Bench is rated high readiness, an explicit benchmark with a solid pipeline; the two first-author NAS papers, RiPPLE and CoRA-NAS, are medium, and RiPPLE states no code release. Benchmark artifacts currently travel better than method papers; releasing code for the NAS methods would most lift this axis.

Research summary

Yifan Yang is a machine learning researcher at Shanghai Jiao Tong University whose profile, name-resolved from public OpenAlex bibliographic data, centers on two threads: efficient neural architecture search (NAS) and multimodal user-intent understanding. In the NAS thread, Yang is first author of RiPPLE (2026), which claims whole-space architecture ranking from early training signals, and CoRA-NAS (2026), a label-free two-stage coarse-ranking and anchor-residual refinement method evaluated on four vision benchmarks. In the multimodal thread, Yang contributed to ODU-Bench (2026), a benchmark for contextual user-demand and intent inference in speech and audio-visual interaction, where Evolve rates reproducibility high. The OpenAlex snapshot shows an exceptionally active 2026 with 30 works, 25 distinct co-authors, and a field distribution of roughly two-thirds NAS to one-third multimodal speech/audio-visual learning; topic counts span five clusters led by multimodal applications (16) and GAN-based image synthesis (8). Evolve's radar scores activity at 100 and reproducibility at 70 (mean over three reviewed papers), while focus is low (8), consistent with the broad spread, and output and impact remain unscored for lack of citation data. Catalog evidence is strong on method-building, with reproducibility caveats: RiPPLE states no code release, and its head-to-head wins depend on proxy features.

Collaboration pattern

The snapshot lists 25 distinct co-authors and a recurring pool of 12 frequent collaborators, including Zhaoyan Wang, Zheng Gao, Xiaoyu Li, Jiaojiao Jiang, Qi Chen, Yunfei Chu, Haolin He, Zihan Liu, Yuxuan Wang, Ziyang Ma, and Ruiyang Xu. Work is team-based: first author on both NAS papers, middle author on the ODU-Bench benchmark, suggesting a method-lead role inside a larger lab cluster spanning architecture search and multimodal interaction.

Research keywords

neural architecture searchzero-cost proxiesperformance predictionlearning curve extrapolationrank aggregationlabel-free NASmultimodal LLMintent inferenceaudio-visual interactionbenchmark designdemand understanding

Highlights

RiPPLE: ranking whole NAS spaces from early training
First-author work claiming whole-space architecture ranking at a fractional training budget; Evolve finds results strong across 12 cells, though head-to-head wins hinge on proxy features and NAS-Bench-201 results land mid-tier, with no code stated.
CoRA-NAS: label-free two-stage NAS ranking
First-author method combining coarse ranking with anchor-residual refinement, claiming cross-space robustness on four vision benchmarks without labels; Evolve judges the evidence largely supportive, though size-space ties the best proxies.
ODU-Bench: benchmarking contextual demand understanding
Middle-author contribution to a benchmark making demand detection and contextual intent inference explicit in multimodal interaction; Evolve rates the pipeline solid and reproducibility high, but notes an LLM-judged metric carries 60% of the score.

Timeline

2026First-author CoRA-NAS introduces label-free coarse ranking with anchor-residual refinement for neural architecture search.
2026First-author RiPPLE claims cross-space NAS performance prediction from early training signals.
2026Contributes to ODU-Bench, a benchmark for contextual user-intent inference in multimodal interaction.
2026OpenAlex attributes 30 works and 25 distinct co-authors to this identity in 2026.

Papers in the Evolve catalog

2026-09-18
Omni Demand Understanding: A Benchmark for Contextual User-Intent Inference in Multimodal Interaction
2026-09-11first author
RiPPLE: Cross-Space Performance Prediction from Early Training for Neural Architecture Search
2026-09-10first author
CoRA-NAS: Coarse Ranking and Anchor-Residual Refinement for Neural Architecture Search

Frequent co-authors

Zhaoyan Wang×2 Zheng Gao×2 Xiaoyu Li×2 Jiaojiao Jiang×2 Qi Chen×1 Yunfei Chu×1 Haolin He×1 Yifan Yang×1 Zihan Liu×1 Yuxuan Wang×1 Ziyang Ma×1 Ruiyang Xu×1

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; citation-linked metrics (h-index, i10-index) are zero or unavailable, so the output and impact axes are unscored.
Yifan Yang is a common name; the profile is name-resolved from public bibliographic data, and the 30-work 2026 count and broad topic spread may include name-ambiguity noise.
The source record lists country DE alongside Shanghai Jiao Tong University; the affiliation-country field is internally inconsistent and should be verified.
Only three papers have been independently reviewed by Evolve (all 2026 arXiv preprints); venue acceptance and peer-review status are not reflected.