Researcher at the Chinese University of Hong Kong working on audio-visual dialogue and efficient LLM reasoning.
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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—
Not yet scored: the bibliometric snapshot contains no publication counts, venue list, or preprint-versus-refereed split. This axis would measure the volume, consistency, and form of the researcher's outputs across the record.
Impact—
Not yet scored: no citation counts, citation windows, or downstream-adoption signals are present in the data. This axis would measure how far the two reviewed 2026 papers and any other outputs have influenced subsequent work.
Focus0/100
The focus score is 0, and the raw field distribution explains it: output splits exactly 0.5 to multimodal machine learning (speech and vision) and 0.5 to machine learning (efficient LLM inference and reasoning). With one 2026 paper in each of two distinct fields, no dominant line has emerged yet, so the score measures dispersion across directions rather than absence of direction.
Collaboration63/100
Collaboration scores 63, backed by raw linkage: 23 distinct coauthors appear in the record, and 12 recur frequently, including Haolin He, Yunfei Chu, Dongchao Yang, and Chunyat Wu. Both reviewed 2026 papers list a middle author position with no corresponding role, so the pattern reflects wide participation in team-scale projects rather than confirmed leadership of them.
Activity—
Not yet scored: yearly publication volume and recency patterns cannot be computed from the snapshot, which only anchors two 2026 papers. This axis would measure publication cadence and sustained engagement over successive years.
Reproducibility55/100
Reproducibility averages 55 across the 2 papers Evolve has reviewed, each rated medium readiness. Rebuilding OmniVChat means re-running a multi-agent synthetic data engine, a rubric benchmark, and RL training; RecurTrace specifies loop-time memory and oracle-distilled halting concretely, though the verdict notes the adaptive margin leans on a floor. The mid-range score reflects real but not turnkey artifacts.
Research summary
Haoning Xu is a researcher at the Chinese University of Hong Kong whose public bibliographic record, resolved under ORCID 0009-0004-7168-7062, spans two equally weighted directions: multimodal machine learning for speech and vision, and machine learning for efficient LLM inference and reasoning. The Evolve catalog holds two reviewed 2026 arXiv papers. OmniVChat builds a multi-agent synthetic data engine, a rubric benchmark, and reinforcement-learning training for native audio-visual dialogue with omni-modal LLMs; Evolve credits gains that transfer to real recordings while flagging judge bias and remaining trade-offs. RecurTrace introduces loop-time memory and oracle-distilled halting for looped transformers, with credible gains over fixed depth on MathQA, though the adaptive margin leans on a floor. Both papers list a middle author position without corresponding status, within a network of 23 distinct coauthors of whom 12 recur frequently. Reproducibility averages 55 across the two papers, both rated medium readiness. Output, impact, and activity are unscored because the snapshot lacks those measures; the focus score of 0 reflects the even 0.5/0.5 field split rather than an absence of direction. This is an early-stage, two-paper profile of an active contributor, not a settled bibliographic record.
Collaboration pattern
Haoning Xu's recorded network spans 23 distinct coauthors, with 12 recurring collaborators including Haolin He, Yunfei Chu, Qi Chen, Dongchao Yang, and Chunyat Wu. Both reviewed 2026 papers place the researcher in a middle author position without corresponding status, so the pattern reads as a contributing member of larger multimodal and LLM-efficiency teams rather than a confirmed lead; the collaboration dimension scores 63 on this basis.
Research keywords
Highlights
Synthetic data engine and RL training for native audio-visual dialogue
Loop-time memory for adaptive latent reasoning in looped LLMs
Timeline
- 2026OmniVChat posted on arXiv: synthesizing, benchmarking, and training for native audio-visual dialogue, in the multimodal speech-and-vision field.
- 2026RecurTrace posted on arXiv: adaptive latent reasoning with loop-time memory for looped LLMs, in the efficient inference and reasoning field.
- 20262026 snapshot: the two reviewed papers split evenly 0.5/0.5 across the two fields, with 23 distinct coauthors recorded and both rated medium reproducibility readiness.
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.
- Output, impact, and activity dimensions are unscored because the bibliometric snapshot contains no publication-count, citation, or yearly-volume measures for this researcher.
- The evidence base is two 2026 arXiv preprints reviewed by Evolve; peer-review status and venue acceptance are not recorded.
- Both catalogued papers list a middle, non-corresponding author position, so individual contribution boundaries within the teams are not characterized.
- The profile is name-resolved from public bibliographic data via ORCID 0009-0004-7168-7062; affiliation country is not recorded — verify identity before citing.