Wen Huang

Johns Hopkins University

Unclaimed · AI-structured profileOpenAlex

Multimodal ML researcher at Johns Hopkins: audio-visual dialogue, demand understanding, and omni-modal LLM benchmarks

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 current data lacks a verified publication count, venue mix, and yearly volume. This axis would measure total research output and its distribution over venues and years.

Impact—

Impact is not scored: no citation counts, h-index, or downstream-usage evidence is present in the data. This axis would measure citation-based influence, such as citations per paper and adoption of the benchmarks.

Focus0/100

Focus scores 0 on the 0-100 scale, grounded in the raw field distribution: the catalogued work splits evenly at 0.5/0.5 between speech-and-audio-visual interaction and speech-and-vision. In practice this reads as a coherent spread across two adjacent multimodal ML sub-areas — audio-visual dialogue systems and omni-modal benchmarking — rather than concentration in a single narrow topic.

Collaboration70/100

Collaboration scores 70. The raw data records 32 distinct coauthors, with recurring partners including Qi Chen, Yunfei Chu, Haolin He, Qiuqiang Kong, Jin Xu, Xie Chen, Yifan Yang, Zihan Liu, Yuxuan Wang, Ziyang Ma, and Ruiyang Xu. Both catalogued 2026 papers are large-team efforts where Wen Huang sits in a middle author position, consistent with a contributor role inside sizeable audio and speech-modeling groups.

Activity—

Activity is not scored: no verified publication timeline or per-year output volume is available beyond the two 2026 catalogued papers. This axis would measure recent publishing cadence, such as papers per year over the last few years.

Reproducibility78/100

Reproducibility scores 78, the mean across the 2 papers reviewed by Evolve. Omni Demand Understanding is rated high readiness on a solid pipeline, while OmniVChat is rated medium; Evolve notes an LLM-judged metric carrying 60% of the ODU-Bench score and judge-bias caveats in OmniVChat as reproducibility-relevant trade-offs.

Research summary

Wen Huang is a researcher affiliated with Johns Hopkins University working in multimodal machine learning. The available bibliometric record splits evenly between two sub-areas — multimodal machine learning for speech and audio-visual interaction (0.5) and for speech and vision (0.5) — and the profile is name-resolved from public bibliographic data. Two 2026 arXiv papers appear in the Evolve catalog, both with Wen Huang in a middle author position. OmniVChat builds a synthetic-data engine, a rubric benchmark, and reinforcement-learning training for native audio-visual dialogue; Evolve's verdict is that gains transfer to real recordings, while judge bias and trade-offs remain. Omni Demand Understanding (ODU-Bench) makes demand detection and contextual intent inference an explicit benchmark for multimodal interaction; its pipeline is rated solid, with the caveat that an LLM-judged metric carries 60% of the score. The collaboration dimension scores 70, based on 32 distinct coauthors, with recurring partners including Qi Chen, Yunfei Chu, Haolin He, Qiuqiang Kong, Jin Xu, Xie Chen, Yifan Yang, Zihan Liu, Yuxuan Wang, Ziyang Ma, and Ruiyang Xu. Reproducibility scores 78 as the mean across the 2 reviewed papers — one high and one medium readiness. Output, impact, and activity are not scored for lack of data.

Collaboration pattern

Wen Huang's bibliometric footprint shows 32 distinct coauthors, with a recurring circle including Qi Chen, Yunfei Chu, Haolin He, Qiuqiang Kong, Jin Xu, Xie Chen, Yifan Yang, Zihan Liu, Yuxuan Wang, Ziyang Ma, and Ruiyang Xu. Both catalogued papers are multi-author, middle-position contributions, suggesting work within larger audio-visual and speech-modeling teams rather than small partnerships; no first or corresponding authorship appears in the current catalog.

Research keywords

audio-visual dialogueomni-modal LLMssynthetic data generationmulti-agent data enginereinforcement learningrubric-based rewardbenchmarkmultimodal LLMdemand understandingintent inferenceaudio-visual interactionfalse trigger

Highlights

OmniVChat: end-to-end pipeline for native audio-visual dialogue
Evolve reviews OmniVChat as combining a synthetic data engine, a rubric benchmark, and reinforcement learning with rubric-based rewards to train native audio-visual dialogue. Reported gains transfer to real recordings, though judge bias and trade-offs remain, and reproducibility readiness is rated medium.
ODU-Bench: demand understanding becomes an explicit multimodal benchmark
Omni Demand Understanding (ODU-Bench) formalizes demand detection and contextual intent inference for multimodal interaction, with false triggers among its focal phenomena. Evolve rates the pipeline solid and reproducibility readiness high, while noting an LLM-judged metric carries 60% of the score.
A 2026 dual track: synthetic data generation plus benchmark construction
Across the two catalogued 2026 papers, Wen Huang's work pairs data-engine synthesis and reinforcement-learning training (OmniVChat) with benchmark design for contextual user intent (ODU-Bench), covering the speech-vision and speech-audio-visual sub-areas in equal measure.

Timeline

2026OmniVChat posted to arXiv, introducing a synthetic data engine, a rubric benchmark, and reinforcement-learning training for native audio-visual dialogue, with middle-position authorship.
2026Omni Demand Understanding (ODU-Bench) posted to arXiv, formalizing demand detection and contextual intent inference for multimodal interaction, also with middle-position authorship.
2026Both papers enter the Evolve catalog in the same year; reviewed reproducibility averages 78 across the two papers, with a recorded footprint of 32 distinct coauthors.

Papers in the Evolve catalog

Frequent co-authors

Qi Chen×2 Yunfei Chu×2 Haolin He×2 Qiuqiang Kong×2 Jin Xu×2 Xie Chen×2 Yunfei Chu×2 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.

Profile is name-resolved from public bibliographic data; 'Wen Huang' is a common name and no ORCID is linked, so identity matching is probabilistic.
Only 2 papers, both 2026 arXiv preprints, are reviewed; output, impact, and activity dimensions are unscored due to insufficient data.
No citation counts, h-index, venue acceptance records, or career history are available; none are inferred.
Reproducibility (mean 78) and collaboration (70) scores cover only the reviewed papers and the coauthor graph visible in this snapshot.