Muhammad Shahid Iqbal Malik

—

Unclaimed · AI-structured profileOpenAlex

Explainable-AI researcher: SHAP/LIME rendering for right-to-left languages and explainable hate-speech detection.

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-01) 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: this axis would measure publication volume and productivity — paper counts, venue types, and first/corresponding-author output over time. The current bibliometric snapshot does not contain those counts, so no score can be computed for this researcher yet.

Impact—

Insufficient data: this axis would measure citation-based influence — citation counts, citation velocity, and uptake of the SHAP/LIME rendering work by other teams. No citation metrics are present in this snapshot, so the axis is left unscored rather than estimated.

Focus0/100

This axis scores 100, the maximum concentration: the raw field distribution contains a single field, explainable AI / NLP visualization, and both catalogued papers sit in explainable AI — one on SHAP/LIME rendering for right-to-left scripts, one on a LIME-based hate-speech detector. Every reviewed output serves one coherent question: making post-hoc explanations usable. With only two papers observed, this perfect focus should be read as a narrow, early profile rather than long-term specialization.

Collaboration28/100

The collaboration score is 14, low: the raw record shows exactly one distinct co-author, Rameesha Zia, the recurring partner on the catalogued work. On both papers this researcher sits in the last author position (non-corresponding), consistent with a small two-person team rather than a lab-scale network. With only two papers in scope, the score reflects a thin observed footprint; broader co-authorship may exist but is not captured in this data.

Activity—

Insufficient data: this axis would track research activity over time — publication cadence, recent-year output, and gaps between releases. Only two 2026 papers are in scope, far too thin to profile a publishing rhythm, so no score is given for this axis.

Reproducibility78/100

Reproducibility scores 100, computed as the mean repro-readiness over one reviewed paper — the SHAP/LIME rendering study, whose pipeline is OCR-measured and described by Evolve as sound and well-controlled, reaching the OCR measurement ceiling. The caveat is coverage: only one of the two catalogued papers carries a rating, so the perfect score reflects a single high-quality assessment rather than a track record across many outputs.

Research summary

Muhammad Shahid Iqbal Malik is an NLP researcher whose catalogued work centers on explainable AI — specifically on whether post-hoc explanation methods remain usable outside idealized rendering conditions. The Evolve catalog holds two last-authored 2026 arXiv papers. The first, "When Explanations Cannot Be Read," shows that SHAP and LIME labels become effectively unreadable when rendered for right-to-left languages such as Urdu in Nastaliq script, with character error rates of 0.82-0.98; an OCR-measured raster-based correction reaches the OCR ceiling, though the fix is word-level only and its impact on human readers is untested. The second proposes a hybrid DistilBERT-BiLSTM-attention framework with LIME explanations for binary and multi-class hate speech detection; Evolve's review finds the reported F1 scores strong but internally inconsistent, with leakage risk and no significance tests, and rates reproducibility readiness medium. The bibliometric radar reflects a young, narrow footprint: a single observed field (explainable AI / NLP visualization), one distinct co-author (Rameesha Zia), a collaboration score of 14, and a perfect reproducibility score computed from one rated paper. Output, impact, and activity axes lack data and are unscored. No institution, country, or ORCID is attached; this profile is name-resolved from public bibliographic data, so treat it as an early-stage view of an identifiable research agenda rather than a complete career record.

Collaboration pattern

The observed collaboration footprint is minimal: exactly one distinct co-author, Rameesha Zia, appears across the catalogued record, and the collaboration dimension scores 14. On both 2026 arXiv papers this researcher sits in the last author position — senior-author placement — though neither lists them as corresponding author. The pattern suggests a small, stable two-person collaboration rather than a distributed network; the broader co-authorship picture cannot be assessed from this data.

Research keywords

explainable AI (XAI)SHAPLIMEright-to-left languagesUrduNastaliq scripthate speech detectionDistilBERTBiLSTMattention mechanismNLP visualization

Highlights

Showing SHAP and LIME explanations are unreadable in right-to-left scripts — and fixing them
In a 2026 arXiv study, this last author's team measured how SHAP and LIME render for right-to-left languages such as Urdu in Nastaliq script, finding character error rates of 0.82-0.98 — labels effectively unreadable. An OCR-measured raster-based fix reached the OCR ceiling, though the correction is word-level only and human impact remains untested. Evolve rates the work sound, well-controlled, and highly reproducible.
An explainable DistilBERT-BiLSTM-attention detector for hate speech
A 2026 arXiv paper proposes a hybrid DistilBERT-BiLSTM model with attention and LIME explanations for binary and multi-class hate speech detection. Evolve's review notes that the reported F1 scores are strong but internally inconsistent, with leakage risk and no significance tests — a promising architecture whose results still await verification. Reproducibility readiness is rated medium.

Timeline

2026Publishes (last author, arXiv) "When Explanations Cannot Be Read," quantifying unreadable SHAP/LIME labels for right-to-left languages at character error rates of 0.82-0.98 and demonstrating an OCR-measured raster-based correction.
2026Publishes (last author, arXiv) an explainable DistilBERT-BiLSTM-attention framework for binary and multi-class hate speech detection, with LIME-based explanations.
2026Both papers enter the Evolve catalog: the rendering study is rated sound and highly reproducible, while the hate-speech framework is flagged for inconsistent F1s and leakage risk; the resulting profile shows one field, one co-author, and one rated paper.

Papers in the Evolve catalog

Frequent co-authors

Rameesha Zia×2 Rameesha Zia×1 Muhammad Shahid Iqbal Malik×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.

Bibliometric scores come from the platform's bibliometric snapshot only; no independent citation database was consulted, and output, impact, and activity axes are unscored for lack of data.
No institution, country, or ORCID is attached to this record; the profile is name-resolved from public bibliographic data.
Only two 2026 arXiv papers were reviewed, and repro-readiness is rated for just one of them; all other career facts fall outside the reviewed scope.
The reproducibility score of 100 and focus score of 100 rest on one rated paper and a single-field distribution, so they should not be read as established track records.