UMass Amherst researcher spanning multimodal vision-language reasoning and full-duplex spoken dialogue systems.
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-02) 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—
Insufficient data: no publication counts, venue breakdown, or cumulative volume are available yet. This axis would measure total research output — paper count, venue tiers, and how volume trends over time.
Impact—
Insufficient data: citation counts, h-index, and downstream adoption are not available for this profile. This axis would measure how often and how widely the reviewed work is cited and built upon by others.
Focus0/100
Scores 0/100, directly reflecting the field distribution: the record splits evenly, 50/50, between multimodal vision-language models and full-duplex spoken dialogue systems. With the reviewed work divided equally across two distinct fields, no dominant specialization has emerged yet; the profile reads as exploratory across two research lines rather than concentrated, which the floor-level score captures faithfully.
Collaboration56/100
Scores 56/100, based on 16 distinct coauthors across the reviewed record. Recurring partners include Sunli Chen, Ding Zhong, Ziqiao Ma, Zeyuan Yang, Lie Lu, Joyce Chai, Chuang Gan, Matthew Sun, and Vinay Kothapally, spanning both the vision-language and speech-dialogue lines. The mid-range score reflects a moderate network; Zhang sits in a middle author position on both catalog papers rather than leading the collaboration.
Activity—
Insufficient data: with only two 2026 arXiv entries in the catalog, recent-year cadence cannot be scored. This axis would measure ongoing publication tempo and year-over-year consistency of research activity.
Reproducibility38/100
Scores 38/100, the mean across 2 reviewed papers, and is mixed. The SVGLM paper on renderable-program multimodal reasoning is rated medium readiness, with gains of roughly 6.5-10 weighted points but no stated significance tests or code release. The full-duplex conversation synthesis harness is rated low readiness, with ablations lacking statistics and Moshi transfer experiments without baselines.
Research summary
Hao Zhang is a researcher at the University of Massachusetts Amherst whose reviewed work spans two distinct areas: multimodal vision-language modeling and speech processing for full-duplex spoken dialogue. The Evolve catalog holds two 2026 arXiv papers with Zhang in a middle author position. The first, Multimodal Thinking with Renderable Programs, introduces SVGLM, which tunes vision-language models to sketch renderable SVG overlays while reasoning, improving geometry-task performance by roughly 6.5-10 weighted points over supervised fine-tuning, though without stated significance tests or code release. The second contributes a synthesis harness for intent-labeled naturalistic full-duplex conversations, with convincing learnability results, though ablations lack statistics and transfer to Moshi lacks baselines. Public bibliometrics show 16 distinct coauthors and an even 50/50 field split, yielding low focus, mid-range collaboration (56/100), and a reproducibility of 38/100 averaged across 2 papers. Recurring collaborators include Sunli Chen, Ding Zhong, Ziqiao Ma, Zeyuan Yang, Lie Lu, Joyce Chai, Chuang Gan, Matthew Sun, and Vinay Kothapally. Because Hao Zhang is a common name and no ORCID is provided, this profile is name-resolved from public bibliographic data and covers only the reviewed items; output, impact, and activity axes remain unscored for insufficient data.
Collaboration pattern
The record shows 16 distinct coauthors across the two reviewed papers. Recurring names include Sunli Chen, Ding Zhong, Ziqiao Ma, Zeyuan Yang, Lie Lu, Joyce Chai, Chuang Gan, Matthew Sun, and Vinay Kothapally. The network spans both research lines, with each paper drawing on a distinct collaborator set; Zhang sits in a middle author position on both catalog papers, consistent with a contributor role rather than lead.
Research keywords
Highlights
Renderable SVG sketches as a reasoning medium for VLMs
A synthesis harness for naturalistic full-duplex conversations
Timeline
- 2026Multimodal Thinking with Renderable Programs appears on arXiv; Zhang is a middle author on the SVGLM work in multimodal vision-language modeling.
- 2026A Harness for Synthesizing Diverse Naturalistic Full-Duplex Conversations appears on arXiv; Zhang is again a middle author, in the speech-dialogue line.
- 2026Evolve's reviewed record for Zhang now spans these two 2026 arXiv preprints across two distinct fields, the basis for this early-stage profile.
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.
- Bibliometric scores are computed from public bibliometric data only; output, impact, and activity axes lack sufficient data and are left unscored.
- The profile rests on only 2 reviewed papers, both 2026 arXiv preprints without confirmed peer-reviewed venue acceptance, so conclusions are early-stage.
- "Hao Zhang" is a common name and no ORCID is provided; the profile is name-resolved from public bibliographic data and may merge or split identities.
- "Hao Zhang" itself appears in the frequent-coauthor list, a name-collision artifact suggesting possible identity mixing in the underlying records.