Mohamed Elhoseiny

Lahore University of Management Sciences

Unclaimed · AI-structured profileOpenAlex

Dual-field researcher spanning multidialectal Arabic speech shared tasks and multi-LLM inference systems at LUMS

papers here

reviewed by Evolve

first-author

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—

Not yet scored: the underlying data on total publication volume and per-year output is unavailable in the current snapshot. This axis would measure the breadth and cadence of Mohamed Elhoseiny's published record beyond the two papers reviewed here.

Impact—

Not yet scored: citation counts, h-index, and downstream uptake are absent from the input data. This axis would measure how widely Mohamed Elhoseiny's papers are cited and adopted; no estimate is made here.

Focus0/100

The computed focus score is 0. The field distribution is an even 50/50 split: half of the cataloged output sits in speech processing / computational linguistics (NLP) and half in machine learning — LLM inference systems. That balance signals deliberate breadth across two active communities rather than a single specialization, so the low score should be read as cross-field spread, not inactivity.

Collaboration60/100

Collaboration scores 60 from 19 distinct coauthors, twelve of whom recur. The recurring set — Nizar Habash, Bashar Talafha, AbdelRahim Elmadany, Yannick Estève, Salima Mdhaffar, Fethi Bougares, Hamzah Luqman, Peter Sullivan, Chiyu Zhang, Ahmed Ashraf, Youssef Mohamed, and Haroun Elleuch — reflects team-scale work: both cataloged papers are large multi-author efforts with Mohamed Elhoseiny in a middle, non-corresponding position.

Activity—

Not yet scored: the snapshot provides no per-year publication timeline beyond the two 2026 papers reviewed, so yearly output rhythm cannot be measured. This axis would track publication cadence and sustained engagement over time; the data is insufficient here.

Reproducibility55/100

Reproducibility scores 55, the mean over the 2 cataloged papers, both rated medium repro-readiness. The NADI 2026 shared task is honestly reported with realism-first evaluations but offers thin TTS/SLT evidence and rankings without statistical tests; COMED's broad multi-LLM gains are credible while its sparse-setting wins fall within noise. Neither paper is flagged as low-reproducibility, but neither is fully turnkey either.

Research summary

Mohamed Elhoseiny, affiliated in the snapshot with Lahore University of Management Sciences, is a full-tier profile built on a deliberately two-sided research footprint. The cataloged output splits evenly, 50/50, between speech processing / computational linguistics (NLP) and machine learning for LLM inference systems, and both papers appeared on arXiv in 2026. On the speech side, he co-authored NADI 2026, the second multidialectal Arabic speech processing shared task, which spans 8 subtasks with realism-first evaluations and is judged honestly reported, though its TTS/SLT evidence is thin and rankings lack statistical tests. On the systems side, he co-authored COMED, which positions an anchor-based accept/verify/collaborate post-controller as the missing middle between model routing and full multi-LLM collaboration; Evolve judges its broad gains credible while noting that wins in sparse settings sit within noise. The collaboration graph holds 19 distinct coauthors, twelve of whom recur, a team-scale pattern spanning both fields; in both cataloged papers he sits in a middle, non-corresponding author position. Computed radar scores give collaboration 60 and reproducibility 55 (the mean over 2 papers, each rated medium repro-readiness), while focus is 0, reflecting the even cross-field split. Output, impact, and activity dimensions are not yet scored because citation, volume, and per-year data are missing. The profile is name-resolved from public bibliographic data and remains early-stage in coverage.

Collaboration pattern

With 19 distinct coauthors and 12 recurring ones, Mohamed Elhoseiny works in team-scale settings. The recurring circle includes Nizar Habash, Bashar Talafha, AbdelRahim Elmadany, Fethi Bougares, Yannick Estève, Salima Mdhaffar, Hamzah Luqman, Peter Sullivan, Chiyu Zhang, Ahmed Ashraf, Youssef Mohamed, and Haroun Elleuch — a pool bridging both cataloged lines. In both papers he holds a middle author position without corresponding-author credit.

Research keywords

Arabic dialect identificationautomatic speech recognitionshared task benchmarkslow-resource languagescode-switchingtext-to-speechmulti-LLM inferencemodel routingLLM cascadesselective collaborationself-consistencytest-time compute

Highlights

NADI 2026: eight Arabic speech subtasks under a realism-first shared task
The second multidialectal Arabic speech processing shared task spans 8 subtasks covering dialect identification, ASR, code-switching, and text-to-speech, with realism-first evaluation and honest reporting; TTS/SLT evidence remains thin and rankings lack statistical tests.
COMED: an anchor-based post-controller between routing and collaboration
In multi-LLM inference, COMED adds an accept/verify/collaborate post-controller that sits between pure routing and full collaboration; Evolve judges its broad gains credible, though wins in sparse settings fall within noise.
A rare dual footprint across Arabic speech and LLM inference systems
Within the reviewed catalog, output splits evenly between speech processing / computational linguistics (NLP) and machine learning — LLM inference systems, pairing low-resource dialect technology with test-time compute research.

Timeline

2026Co-authored NADI 2026, the second multidialectal Arabic speech processing shared task spanning 8 subtasks across dialect identification, ASR, code-switching, and text-to-speech.
2026Co-authored COMED, proposing an anchor-based accept/verify/collaborate post-controller positioned between routing and collaboration in multi-LLM inference.
2026Both cataloged papers appear on arXiv in the same year, splitting evenly between the speech processing/NLP and LLM inference systems fields.

Papers in the Evolve catalog

Frequent co-authors

Peter Sullivan×1 Bashar Talafha×1 Ahmed Ashraf×1 Fethi Bougares×1 Haroun Elleuch×1 Chiyu Zhang×1 AbdelRahim Elmadany×1 Youssef Mohamed×1 Salima Mdhaffar×1 Yannick Estève×1 Hamzah Luqman×1 Nizar Habash×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.

Identity is name-resolved from public bibliographic data; no ORCID is linked in the snapshot, so the affiliation (LUMS) may reflect a name match rather than verified authorship.
Citation, h-index, and yearly-output data are absent from the snapshot, so the output, impact, and activity axes are unscored rather than estimated.
Only 2 papers from 2026 are in the reviewed catalog, both arXiv preprints with middle, non-corresponding authorship, so the profile reflects a narrow slice of activity.
Reproducibility (55) and collaboration (60) are means over this small 2-paper sample and may shift as more papers are reviewed.