Yunfei Chu

Alibaba Group (China) · China

Unclaimed · AI-structured profileOpenAlex

Alibaba researcher in omni-modal AI: audio-visual dialogue, embodied spatial-audio benchmarks, synthetic data engines

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 scoring is not yet available: OpenAlex lacks a complete work record for this author beyond the recent window (13 works in three years). This axis would measure total publication volume, venue mix and output trajectory over time.

Impact—

Impact cannot be scored: OpenAlex citation coverage is too thin (h-index 1, i10-index 0) to support a reliable measure. This axis would capture citation counts, citation velocity and influence on subsequent research across the author's papers.

Focus0/100

The focus score is 0, the lowest band, and the field distribution explains why: research splits almost evenly — about 33% each — across speech and audio-visual interaction, speech and vision, and embodied audio-visual AI. Topic counts (multimodal applications 8, speech recognition and synthesis 5, music and audio 5) confirm breadth across adjacent audio-centric areas rather than concentration in one niche.

Collaboration76/100

Collaboration scores 76 on a 0-100 scale, backed by 44 distinct co-authors within the recent three-year window. All three reviewed papers place this researcher in middle authorship without a corresponding-author role, consistent with contributory positions on large industrial teams; recurring collaborators include Qi Chen, Jin Xu, Xie Chen, Yuxuan Wang and Ziyang Ma.

Activity77/100

Activity scores 77, and the raw counts show a steep recent ramp: OpenAlex logs 13 works in the past three years, with counts by year rising from a single work in 2024 to 12 works in 2026. All three catalog papers are 2026 arXiv preprints, indicating a concentrated, ongoing burst of output rather than long-term accumulation.

Reproducibility58/100

Reproducibility scores 58, computed by Evolve across 3 reviewed papers (reproMean 58). Readiness varies widely: ODU-Bench is rated high, OmniVChat medium, and OmniEcho low — the latter's state-of-the-art claims rest on its own benchmark. The pattern suggests real artifact-building capability with uneven support for external reimplementation.

Research summary

Yunfei Chu is a researcher at Alibaba Group (China) working at the intersection of speech, audio, vision and language — an emerging profile built around omni-modal foundation models. OpenAlex records 13 works in the past three years, 12 of them dated 2026, an activity score of 77 and a collaboration score of 76 with 44 distinct co-authors, indicating fast-moving, heavily team-based industrial research. Topic records concentrate on multimodal machine learning applications (8), speech recognition and synthesis (5), music and audio processing (5) and speech and dialogue systems (4). The reviewed portfolio spans three evenly weighted subfields: speech and audio-visual interaction, speech and vision, and embodied audio-visual AI. Three 2026 arXiv papers anchor the catalog: OmniEcho introduces a real-world first-order ambisonics benchmark (2,972 QA pairs and 900 navigation tasks) for omni-modal embodied agents; OmniVChat combines a synthetic data engine, a rubric benchmark and reinforcement learning for native audio-visual dialogue; and ODU-Bench formalizes demand detection and contextual intent inference, with high reproduction readiness. Evolve's verdicts are balanced: gains are real but benchmarks are largely self-defined, and LLM-judged metrics carry substantial weight. Citation-level evidence remains thin — h-index 1 and i10-index 0 in OpenAlex — so impact cannot yet be scored.

Collaboration pattern

Collaboration is intense and team-based: 44 distinct co-authors in the recent three-year window underpin a collaboration score of 76. All three reviewed 2026 papers list this researcher in middle authorship with no corresponding-author role, suggesting contributory positions on large Alibaba project teams. Recurring co-author names include Qi Chen, Jin Xu, Xie Chen, Yuxuan Wang and Ziyang Ma; note that "Yunfei Chu" itself appears in the co-author list, hinting at same-name ambiguity in OpenAlex resolution.

Research keywords

audio-visual dialogueomni-modal large language modelsmultimodal LLM benchmarkssynthetic data generationreinforcement learning with rubric rewardsspeech recognition and synthesisspatial audio (first-order ambisonics)audio-visual navigationembodied AI agentsintent inferencespeech and dialogue systemsmultimodal machine learning

Highlights

OmniEcho: a real-world FOA benchmark for omni-modal embodied agents
Introduces a real-world first-order ambisonics (FOA) benchmark with 2,972 QA pairs and 900 navigation tasks, plus an FOA-augmented omni model; Evolve finds the gains real but modest, with SOTA claims resting on its own benchmark (reproducibility: low).
OmniVChat: synthetic data engine and RL for native audio-visual dialogue
Combines a multi-agent synthetic data engine, a rubric-based benchmark and reinforcement learning with rubric rewards for native audio-visual dialogue; gains transfer to real recordings, though judge bias and performance trade-offs remain (reproducibility: medium).
ODU-Bench makes contextual user-demand inference measurable
Turns demand detection and contextual intent inference in multimodal interaction into an explicit benchmark; the pipeline is solid and reproducibility is rated high, but an LLM-judged metric carries 60% of the score.

Timeline

2024First recorded OpenAlex output: 1 work indexed for this author in 2024.
2026Output accelerates sharply: 12 works indexed for 2026, within a recent three-year total of 13.
2026Contributes to OmniEcho, a real-world FOA benchmark (2,972 QA, 900 navigation) and FOA-augmented omni model for embodied spatial audio understanding (middle authorship).
2026Contributes to OmniVChat, pairing a synthetic data engine and rubric benchmark with reinforcement learning for native audio-visual dialogue (middle authorship).
2026Contributes to ODU-Bench, a benchmark for demand detection and contextual intent inference in multimodal interaction, with high reproduction readiness (middle authorship).

Papers in the Evolve catalog

Frequent co-authors

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

All bibliometrics come from a single OpenAlex snapshot; h-index 1 and i10-index 0 reflect that index's thin coverage for this author, not a verified influence estimate.
No ORCID is linked and identity is name-resolved via OpenAlex bibliographic data; the name also appears inside its own co-author list, so same-name merging or splitting is possible.
Output and impact axes are unscored because the underlying citation and work-count data is insufficient (radar.dimMissing).
All reviewed papers are 2026 arXiv preprints with middle authorship and no corresponding-author role; no peer-reviewed venue records appear in the provided data.