Xingyu Shen

—

Unclaimed · AI-structured profileOpenAlex

Early-stage researcher spanning AI-generated image detection datasets and robust ASR front-end design

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

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—

Data not yet available: publication counts, venue mix, and per-year output are absent from the indexed record, which covers two 2026 arXiv preprints. This axis would measure total research production and its distribution across venues and years.

Impact—

Data not yet available: citation counts and downstream uptake are absent from the indexed record. This axis would measure the influence of the work on subsequent research, practice, and detector or benchmark adoption.

Focus100/100

Scores 100 for topical concentration. The only field indexed in the profile's distribution is computer vision (AI-generated image detection), and the reviewed dataset paper sits squarely in detector benchmarking, generator attribution, and C2PA provenance. Note the asymmetry: only one paper is field-classified, and the other 2026 preprint targets ASR front-ends, so this maximal concentration reflects a thin record and should be re-read as the field distribution fills in.

Collaboration50/100

Scores 50 on a footprint of 11 distinct co-authors recorded across the two indexed 2026 papers. The recurring co-author list includes Wei-Ping Zhu, Benoit Champagne, Dennis Ng, Ankit Raj, and Yuchen Zhou, among others. Shen appears as first author on the ASR front-end preprint and as a middle author on the dataset study, consistent with multi-author team research rather than solo output.

Activity—

Data not yet available: recent-year publication cadence cannot be computed from a snapshot of two 2026 arXiv preprints. This axis would measure ongoing research tempo and year-over-year continuity of output.

Reproducibility55/100

Scores 55, computed as reproMean 55 over a single assessed paper (reproPapers 1). The ChatGPT Images 2.5 dataset study carries a medium reproducibility-readiness rating; Evolve's verdict ties this to transparent launch-window dataset construction and candid attribution limits, while noting weak detector results and exploratory cross-cohort contrasts. With one paper assessed, this score should be treated as an early, low-confidence signal.

Research summary

Xingyu Shen is an early-stage researcher whose indexed output to date consists of two 2026 arXiv preprints spanning two applied machine-learning areas. The first, 'Parallel Time-Band Mixing with Learned Observation-Adding for Robust ASR Front-Ends,' is a first-author contribution to robust automatic speech recognition front-end design. The second, 'ChatGPT Images 2.5 in the Wild: A Launch-Period Dataset and Detector Evaluation,' places Shen in a middle author position on a dataset and benchmarking effort in AI-generated image detection; Evolve's review calls it a transparent launch-window dataset with candid attribution limits — an honest dataset paper whose detector results are weak and cross-cohort contrasts exploratory — and rates its reproducibility readiness as medium. Focus scores 100, reflecting the single indexed field (computer vision, AI-generated image detection); collaboration scores 50 across 11 distinct co-authors, including recurring collaborators Wei-Ping Zhu and Benoit Champagne; reproducibility scores 55 from one assessed paper. Output, impact, and activity dimensions are unscored because citation and production-volume data are absent. No institution, country, or ORCID is on record, the name is resolved from public bibliographic data, and the profile should be read as provisional until a larger record accrues.

Collaboration pattern

The collaboration dimension of 50 reflects a team-based working pattern: 11 distinct co-authors appear across the two indexed 2026 papers. The recurring list includes Dennis Ng, Ankit Raj, Kidus Zewde, Tommy Duong, Yuchen Zhou, Yuxin Zhang, Neo Tiangratanakul, Simiao Ren, Runze Wang, Wei-Ping Zhu, and Benoit Champagne. Shen takes a first-author role on the ASR front-end paper and a middle-author role on the larger dataset effort, consistent with contributing within multi-author groups.

Research keywords

AI-generated image detectionChatGPT Images 2.5dataset curationgenerator attributionC2PA provenancedetector benchmarkingASR front-endstime-band mixinglearned observation-addingrobust speech recognitionaudio signal processing

Highlights

Launch-window dataset for detecting ChatGPT Images 2.5 output
As a middle author, Shen contributed to a 2026 dataset and detector-evaluation study built around the ChatGPT Images 2.5 launch window. Evolve's verdict credits a transparent dataset with candid generator-attribution limits, while flagging weak detector results and exploratory cross-cohort contrasts; reproducibility readiness is rated medium.
First-author work on robust ASR front-ends
Shen leads 'Parallel Time-Band Mixing with Learned Observation-Adding for Robust ASR Front-Ends' (2026, arXiv) as first author, proposing parallel time-band mixing with a learned observation-adding mechanism for more robust speech-recognition front-ends. Evolve has not yet published a review verdict or reproducibility assessment for this paper.
Reproducibility posture grounded in one assessed paper
The profile's reproducibility dimension of 55 rests on a single assessed paper — the ChatGPT Images 2.5 dataset study, whose medium readiness reflects open dataset construction and candid attribution limits rather than validated detector performance.

Timeline

2026First-author arXiv preprint 'Parallel Time-Band Mixing with Learned Observation-Adding for Robust ASR Front-Ends' appears, contributing a robust front-end design for automatic speech recognition.
2026Co-authors 'ChatGPT Images 2.5 in the Wild: A Launch-Period Dataset and Detector Evaluation' in a middle author position, a dataset and detector-benchmarking study in AI-generated image detection.
2026The dataset paper enters the Evolve catalog with a review verdict — transparent but with weak detector results — and a medium reproducibility-readiness rating.
2026The indexed collaboration footprint reaches 11 distinct co-authors across the two 2026 preprints, with recurring collaborators including Wei-Ping Zhu and Benoit Champagne.

Papers in the Evolve catalog

2026-09-14
ChatGPT Images 2.5 in the Wild: A Launch-Period Dataset and Detector Evaluation

Frequent co-authors

Dennis Ng×1 Ankit Raj×1 Kidus Zewde×1 Tommy Duong×1 Yuchen Zhou×1 Yuxin Zhang×1 Neo Tiangratanakul×1 Simiao Ren×1 Runze Wang×1 Wei-Ping Zhu×1 Benoit Champagne×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.

Bibliometric inputs derive from OpenAlex-based public bibliographic data; institution, country, and ORCID are not on record for this profile.
Only two 2026 arXiv preprints are in scope, so output, impact, and activity dimensions remain unscored and career-stage readings are provisional.
The ASR front-end paper has not been reviewed by Evolve; its field classification, keywords, and reproducibility readiness are unclassified.
The name Xingyu Shen is resolved from public bibliographic data and may correspond to more than one researcher; verify via the listed co-authors before citing.