Shanghai Jiao Tong University researcher building benchmarks and RL training for speech-centric audio-visual dialogue.
2
papers here
2
reviewed by Evolve
0
first-author
0
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—
No publication-volume data is available yet: this axis would measure total output — papers per year, venue tier, and first/corresponding-author share. The catalog currently holds only 2 reviewed 2026 papers, too few to score.
Impact—
Impact cannot be scored yet: it would require citation counts, citing works, and h-index trajectories over time, none of which are in the current data. The two 2026 preprints are too recent for meaningful citation traces.
Focus0/100
Radar records a focus score of 0, yet the raw field distribution behind it is narrow: cataloged work splits evenly (0.5 and 0.5) between two adjacent, speech-centered subfields — multimodal learning for speech and audio-visual interaction, and for speech and vision. With only two papers in the catalog, this axis is under-determined; the distribution itself points to a tightly concentrated speech-centric agenda rather than scattered topics.
Collaboration70/100
The raw record lists 32 distinct coauthors, with 11 names recurring — Qi Chen, Yunfei Chu, Haolin He, Wen Huang, Qiuqiang Kong, Jin Xu, Yifan Yang, Zihan Liu, Yuxuan Wang, Ziyang Ma, Ruiyang Xu — supporting a radar collaboration score of 70. Both 2026 papers are large-team outputs, with Xie Chen mid-author on OmniVChat and last author on ODU-Bench, indicating contribution inside sizeable multimodal-speech groups rather than solo work.
Activity—
Activity data is insufficient: this axis would track annual publication cadence and ongoing output rhythm. Only two 2026 papers are on record, with no year-by-year history, so no trend can be computed.
Reproducibility78/100
Evolve's reproducibility assessment covers 2 papers with a mean readiness score of 78. ODU-Bench rates high — a benchmark pipeline built to be rerun — while OmniVChat rates medium, its synthetic data engine and reinforcement-learning training adding heavier reproduction cost. The 78 average reflects strength in reusable benchmark assets alongside training-side components that remain harder to replicate.
Research summary
Xie Chen is a multimodal machine learning researcher at Shanghai Jiao Tong University whose cataloged work centers on speech-centric audio-visual interaction. Evolve's catalog holds two 2026 arXiv papers. OmniVChat (middle author) builds a full stack for native audio-visual dialogue — a multi-agent synthetic data engine, a rubric-based benchmark, and reinforcement learning with rubric-based rewards — with Evolve noting that gains transfer to real recordings while judge bias and trade-offs remain. ODU-Bench (last author) turns demand detection and contextual user-intent inference into an explicit benchmark spanning audio-visual interaction and false-trigger cases; its pipeline is solid, though an LLM-judged metric carries 60% of the score. The field distribution splits evenly (0.5 and 0.5) between speech-and-audio-visual and speech-and-vision multimodal learning, indicating a tightly focused agenda. Bibliometrics show 32 distinct coauthors, a collaboration score of 70, and 11 recurring collaborators; reproducibility averages 78 across 2 assessed papers. Together, the two papers sketch a benchmark-plus-training agenda for omni-modal dialogue systems. Output, impact, and activity scores are not yet computable.
Collaboration pattern
With 32 distinct coauthors and 11 recurring collaborators — including Qi Chen, Yunfei Chu, Jin Xu, Ziyang Ma, and Qiuqiang Kong — Xie Chen works within sizeable, speech-centric multimodal teams. Authorship placement (middle on OmniVChat, last on ODU-Bench, neither corresponding) suggests a senior-but-supporting role in group projects spanning data engines, benchmarks, and RL training. The radar collaboration score of 70 reflects broad, repeated co-authorship rather than isolated work.
Research keywords
Highlights
A full-stack recipe for native audio-visual dialogue (OmniVChat)
Making demand understanding a measurable benchmark (ODU-Bench)
Reproducibility-aware benchmark practice
Across the 2 papers with readiness ratings, reproducibility averages 78: ODU-Bench rates high and OmniVChat medium — benchmark assets are built to be rerun, while training-side components stay heavier to replicate.
Timeline
- 2026OmniVChat appears on arXiv (middle author): a synthetic-data, rubric-benchmark, and reinforcement-learning recipe for native audio-visual dialogue.
- 2026ODU-Bench is published on arXiv (last author): demand detection and contextual user-intent inference become an explicit multimodal benchmark.
- 2026Both papers land in the same year, marking a concentrated speech-centric audio-visual agenda spanning data engines, benchmarks, and RL training.
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 and radar scores are derived from public bibliographic data (OpenAlex-style coverage); citation and venue coverage may be incomplete.
- Xie Chen is a common romanized name; this profile is name-resolved from public bibliographic data and the two listed papers, and may overlap with other researchers sharing the name.
- No ORCID, country, publication counts, citation totals, or h-index are provided; output, impact, and activity axes remain unscored.
- Both cataloged papers are 2026 arXiv preprints; peer-reviewed venue status is not recorded.