University of Auckland researcher in audio-language models and perception-grounded test-time reinforcement learning
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
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—
Not yet scored: this profile shows h-index 0, i10-index 0 and only one year (2026) of per-year counts, so no cumulative output volume is available to measure. This axis would track total and per-year publications once a longer indexed record exists.
Impact—
Not yet scored: no citation counts or citation-based influence data are available (h-index and i10-index both 0), so impact cannot be estimated. This axis would measure citation uptake and scholarly influence of the work once citing literature accumulates.
Focus100/100
Focus scores 100, reflecting a highly concentrated topic profile: OpenAlex topic counts are led by Speech Recognition and Synthesis (5) and Multimodal Machine Learning Applications (5), with Speech and Audio Processing (3), Topic Modeling (2) and Music and Audio Processing (2) nearby; the cataloged field distribution lists a single field, multimodal machine learning (audio-language models). Work clusters tightly around audio, speech and audio-language modeling.
Collaboration50/100
Collaboration scores 50, a mid-range value: the OpenAlex record lists 11 distinct coauthors across the recent three-year window, and the network includes 11 frequent collaborators such as Ting Dang, Xiaofeng Yu and Jean Honorio. Both cataloged 2026 arXiv papers were written from middle-author positions rather than solo work, indicating steady but not yet dense co-authorship.
Activity72/100
Activity scores 72: OpenAlex counts 11 works in 2026, and the recent three-year window also totals 11 works, making 2026 the only recorded publication year so far. Both cataloged papers appeared on arXiv in 2026 within audio-language and multimodal topics, indicating a burst of recent output rather than a long-running publication cadence.
Reproducibility55/100
Reproducibility scores 55 (reproMean 55), but the base is thin: only one cataloged paper has been assessed (reproPapers 1). That paper, the PG-TTRL test-time reinforcement learning work, is rated medium reproducibility readiness; Evolve notes consistent Avg@8 and Pass@k gains alongside thin statistics, so the score should be read as an early, single-paper signal.
Research summary
Hong Jia is a researcher at the University of Auckland (New Zealand) whose profile, resolved from public OpenAlex bibliographic data, centers on audio, speech and multimodal machine learning. OpenAlex topic counts place Speech Recognition and Synthesis (5) and Multimodal Machine Learning Applications (5) at the top, followed by Speech and Audio Processing (3), Topic Modeling (2) and Music and Audio Processing (2) — a tightly concentrated record yielding a focus score of 100. The indexed record is young: h-index and i10-index are both 0, with 11 works in the recent three-year window and 11 recorded in 2026, giving an activity score of 72. Two 2026 arXiv papers anchor the catalog. 'Listen Then Reason' proposes PG-TTRL, a perception-grounded test-time reinforcement-learning method for large audio-language models built on Qwen2.5-Omni; Evolve finds consistent Avg@8 and Pass@k gains but notes Maj@8 often trails baselines with thin statistics, rating reproducibility readiness medium (reproMean 55 from one assessed paper). 'ManGo: Manga Active Narrative Grounding Optimization' extends grounding to visual manga narratives. Collaboration is mid-range (score 50), with 11 distinct coauthors and frequent partners including Ting Dang, Xiaofeng Yu and Jean Honorio; both cataloged papers list this author in a middle position rather than first or corresponding.
Collaboration pattern
The OpenAlex record lists 11 distinct coauthors over the recent three-year window, matching a collaboration score of 50 — active but not yet dense. The network highlights 11 frequent collaborators, including Ting Dang, Xiaofeng Yu, Jean Honorio, Abhirup Ghosh and Jiaheng Dong. Both 2026 cataloged arXiv papers were written from middle-author positions, consistent with team-based deep-learning work rather than solo authorship; no corresponding-author role appears in the cataloged record.
Research keywords
Highlights
Perception-grounded test-time RL for audio-language models
Honest evaluation trade-offs on majority voting
Extending grounding to manga narratives
'ManGo: Manga Active Narrative Grounding Optimization' (2026, arXiv) extends the grounding agenda from audio to manga visual-narrative optimization; the paper is listed in the catalog without a full Evolve review yet.
Timeline
- 2026Publishes 'Listen Then Reason' on arXiv (middle author): PG-TTRL weights test-time reinforcement-learning updates by acoustic grounding for large audio-language models.
- 2026Co-authors 'ManGo: Manga Active Narrative Grounding Optimization' on arXiv, extending grounding from audio to manga visual narratives.
- 2026OpenAlex records 11 works in 2026 (the recent three-year total is also 11) with 11 distinct coauthors — the only recorded publication year so far.
- 2026Evolve assesses the PG-TTRL paper's reproducibility as medium readiness, contributing a reproMean of 55 from one assessed paper.
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 come from an OpenAlex author profile only (no ORCID supplied); h-index and i10-index are recorded as 0, which may reflect a young or fragmented profile rather than actual impact.
- 'Hong Jia' is name-resolved from public OpenAlex bibliographic data; without an ORCID, identity cannot be fully disambiguated from same-named researchers.
- Only 2 papers by this author are in the Evolve catalog, and the reproducibility score rests on a single assessed paper (reproMean 55, reproPapers 1).
- The output and impact radar axes are unscored because citation and cumulative per-year output data are insufficient in the current snapshot.