Early-career AI researcher at Western Michigan University spanning retrieval-augmented generation and computer vision.
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
0
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: the record is too young for a stable output measure. The only signals are h-index 0, i10-index 0, and 6 works all dated 2026. This axis would measure publication volume, venue quality, and completeness over time.
Impact—
Not yet scored: no citation data is available and both h-index and i10-index read 0. This axis would measure citations, uptake, and downstream influence of the work; with all outputs dated 2026, impact cannot yet be meaningfully assessed.
Focus0/100
Scores 0/100, reflecting an even split in fieldDist: 0.5 for computer vision (artificial intelligence) and 0.5 for natural language processing / retrieval-augmented generation. Topic records confirm the breadth — AI in healthcare and education (3), advanced neural network applications (2), multimodal machine learning applications (2), machine learning in healthcare (2), and topic modeling (1). No single field yet dominates the publication record.
Collaboration44/100
Scores 44/100. Bibliographic records list 8 distinct coauthors, with recurring partners such as Bahar Uddin Mahmud, Ajay Gupta, Guan Yue Hong, Hexu Liu, Halil Dursunoglu, Charles Rodgers, and Alvis Fong. The pattern suggests small, largely local teams at Western Michigan University rather than large distributed consortia; the moderate score reflects a network that is real but still narrow and early in formation.
Activity57/100
Scores 57/100. OpenAlex records 6 works in the most recent three-year window, and the countsByYear data places all of them in 2026 — a burst of very recent output rather than a sustained multi-year record. Two of these works are arXiv preprints reviewed by Evolve. With activity concentrated in a single year, the score should be read as momentum, not an established publishing cadence.
Reproducibility38/100
Scores 38/100, a mean computed over 2 reviewed papers (reproPapers: 2, reproMean: 38). CiteGuard-RAG is rated medium readiness, while the commonsense-in-vision survey is rated low, with Evolve citing unverifiable numbers, broken citations, and PRISMA arithmetic errors. Reproducibility is currently the weakest documented axis, and improvement depends on releasing validation code and checkable survey protocols.
Research summary
Sumit Barua is an early-career artificial-intelligence researcher at Western Michigan University (US) whose documented record, resolved from OpenAlex bibliographic data, sits at the intersection of computer vision and natural language processing — the field distribution is an even 50/50 split. The snapshot records 6 works in the most recent three-year window, all dated 2026, with 8 distinct coauthors and h-index and i10-index both at 0, consistent with a profile at its very beginning. Two arXiv preprints have been reviewed by Evolve. The first, CiteGuard-RAG (2026, first author), presents a validation-centered system for evidence-grounded question answering, adding sentence-level grounding validation, citation-validity checks, and abstention calibration for legal QA; Evolve judges the architecture sensible but flags that the headline 98.3% and 0.0% figures come from the same validation layer, making them close to circular. The second, a co-authored survey of commonsense reasoning in computer vision (2026), covers knowledge graphs, scene graphs, neuro-symbolic models, and vision transformers, though Evolve notes unverifiable numbers, broken citations, and PRISMA arithmetic errors that undercut its systematic claim. Topic records cluster around AI in healthcare and education (3), advanced neural network applications (2), multimodal machine learning (2), machine learning in healthcare (2), and topic modeling (1). Reproducibility averages 38/100 across the two reviewed papers — medium for the system paper, low for the survey — indicating that releasing code and checkable protocols would most strengthen the record.
Collaboration pattern
Records show 8 distinct coauthors. The recurring partners are Bahar Uddin Mahmud, Guan Yue Hong, Ajay Gupta, Hexu Liu, Guan Hong, Halil Dursunoglu, Charles Rodgers, and Alvis Fong — a set consistent with small teams anchored at Western Michigan University. Both reviewed papers are multi-author arXiv preprints spanning NLP/RAG and computer vision, suggesting one loosely coupled local group collaborating across topics rather than separate external partnerships. The collaboration score of 44/100 reflects this narrow but active network.
Research keywords
Highlights
CiteGuard-RAG — first-author validation layer for evidence-grounded QA
Broad survey of commonsense reasoning in computer vision
Timeline
- 2026Posted CiteGuard-RAG on arXiv as first author — a validation-centered system for evidence-grounded question answering.
- 2026Co-authored an arXiv survey on commonsense reasoning in computer vision, covering knowledge graphs, scene graphs, neuro-symbolic models, and vision transformers.
- 2026OpenAlex records 6 works in 2026 — all visible activity in this profile falls within this single year, with h-index and i10-index still at 0.
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 OpenAlex only; h-index and i10-index (both 0) reflect that snapshot, not a full citation audit.
- Only 2 papers have been independently reviewed by Evolve, and both are 2026 arXiv preprints whose peer-review status is not documented here.
- Output and impact radar axes are unscored because citation and venue data are insufficient at this career stage.
- The profile is name-resolved from public bibliographic data; no biography, award, or identity information beyond the provided record is used.