McGill-affiliated researcher working across quantum-augmented diffusion language models and embodied AI benchmarks
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: publication-volume data is insufficient. This axis would measure total output, venue distribution and yearly productivity; the profile currently rests on only two reviewed 2026 arXiv preprints with middle authorship.
Impact—
Not yet scored: citation and influence data are unavailable. This axis would measure citations, downstream adoption and field-wide reach; neither paper's citation record is present in the current bibliometric snapshot.
Focus0/100
Focus scores 0 because Xiong's reviewed output splits evenly — 50% embodied AI/robotics and 50% quantum machine learning (quantum NLP) — with no dominant field yet. The two 2026 papers address unrelated problems: quantum circuit generation for diffusion language models versus physics-grounded benchmark design for embodied multimodal LLMs. This even split signals exploratory breadth rather than an established specialization, so the score reflects distribution balance, not absence of work.
Collaboration48/100
Collaboration scores 48, based on 10 distinct coauthors across the reviewed work. Both papers carry middle author positions without corresponding authorship, suggesting a contributor role rather than team leadership so far. The co-author records show a repeated name — Bang Liu appears twice in the frequent-collaborator list — indicating at least one stable working relationship within a compact but moderately sized network typical of early arXiv-stage output.
Activity—
Not yet scored: longitudinal productivity data is unavailable. This axis would measure publication cadence and sustained engagement over time; the snapshot covers only two 2026 papers, too few to infer any activity trend.
Reproducibility55/100
Reproducibility scores 55, a mean computed over exactly two papers, both rated medium readiness by Evolve. The ReactHuman benchmark is described as reproducible simulated infrastructure, while HyperQ's reported width gains lack error bars and verified baselines. Both works are criticized for leaning on descriptive statistics or unverified comparisons. The score indicates partially reproducible work, with statistical rigor and baseline verification as the main open gaps.
Research summary
Mengyang Xiong is a McGill University researcher whose reviewed output to date sits at the junction of two distinct areas: quantum machine learning for language models, and embodied AI evaluation. The Evolve catalog holds two 2026 arXiv papers, in both of which Xiong appears as a middle author without corresponding authorship. The first, 'Circuit Hypernetworks for Quantum-Augmented Diffusion Language Models,' presents HyperQ, which emits per-token IQP circuits inside a frozen 1.1B-parameter masked diffusion language model; Evolve judges the reported width gains as plausible but resting on only three widths, without error bars and with unverified baselines. The second, 'ReactHuman,' is a physics-grounded simulated benchmark scoring how multimodal LLMs make human-like reactive decisions in safety-critical situations, credited for strong diagnostic design while relying on descriptive statistics and unvalidated metrics. Bibliometric signals rest on a small base: ten distinct coauthors, a 50/50 field split between embodied AI/robotics and quantum machine learning, a collaboration score of 48, and a reproducibility mean of 55 computed over the two papers. Output, impact, and activity metrics are not yet available, making this an early-stage profile built from a narrow slice of public bibliographic data.
Collaboration pattern
The reviewed work shows a moderately connected network: 10 distinct coauthors with a collaboration score of 48. Both papers list Xiong in a middle author position, non-corresponding, consistent with contributing within a larger team. Recurring collaborators include Bang Liu (listed twice among frequent coauthors), alongside Yizhan Li, Jianxin You, Zicheng Zhao, Dekun Wu and others, suggesting overlapping project teams across both the quantum NLP and embodied AI lines of work.
Research keywords
Highlights
HyperQ: per-token IQP circuits inside a frozen 1.1B diffusion language model
ReactHuman: physics-grounded benchmark for reactive hazard decisions in multimodal LLMs
Timeline
- 2026ReactHuman, a physics-grounded benchmark for reactive decision-making in embodied multimodal LLMs, appears on arXiv with Xiong as a middle author.
- 2026Circuit Hypernetworks for Quantum-Augmented Diffusion Language Models appears on arXiv; HyperQ emits per-token IQP circuits in a frozen 1.1B diffusion LM, again with middle authorship.
- 2026Both papers enter the Evolve catalog with medium reproducibility readiness, establishing a profile split evenly between embodied AI and quantum machine learning.
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.
- Identity is name-resolved from public bibliographic data; no ORCID or country field was available for verification.
- All quantitative dimensions rest on only two reviewed 2026 arXiv preprints, so output, impact and activity scores are missing due to insufficient data.
- Venue records, citation counts, h-index and funding information are not part of the input snapshot and are therefore not reported.
- Both papers carry middle, non-corresponding authorship, so specific contributions within each paper cannot be attributed from the available data.