Table-centric document AI researcher at IIT Bombay — multi-script table detection and table structure/QA pipelines
2
papers here
2
reviewed by Evolve
1
first-author
1
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—
Output is not scored: this snapshot contains no verified publication counts, venue totals, or output-volume metrics for Dhruv Kudale. The axis would measure how much research output the researcher produces per year and across publication venues.
Impact—
Impact is not scored: no citation counts, h-index, or citation-window data is present in the input. The axis would measure how often the researcher's work is cited and the influence it exerts on the field.
Focus0/100
The focus score computes to 0 because the reviewed output splits evenly — 0.5 versus 0.5 — across two labeled fields: document intelligence (computer vision / NLP, table understanding) and computer vision for document image analysis. In substance the two labels describe one theme, since both 2026 papers target tables in document images, so the low score reflects labeling breadth more than genuinely scattered research.
Collaboration36/100
Collaboration scores 36 from a pool of just 5 distinct co-authors across the two reviewed papers. The recurring names — Ganesh Ramakrishnan, Jahanvi Rajput, Saikiran Kasturi, Utkarsh Verma and Udhay Brahmi — form one dense IIT Bombay cluster, and Dhruv Kudale is first and corresponding author on EMBLEM. The pattern suggests early research leadership inside a compact team, with little evidence yet of external or cross-institution networks.
Activity—
Activity is not scored: beyond the two 2026 preprints, no per-year publication records are available, so publishing cadence cannot be computed. The axis would measure year-over-year consistency and rhythm of research output over recent years.
Reproducibility100/100
Reproducibility is the standout axis at 100: the reproMean over 2 reviewed papers is 100, and both carry Evolve's 'high' reproducibility readiness. Evolve's verdicts reference concrete, re-runnable artifacts — DELTA's modular TATR+SPRINT+OCR pipeline emitting OTSL, and EMBLEM's new MANDALA benchmark with quantified F1 deltas — supporting the view that results are specified precisely enough to be independently re-tested.
Research summary
Dhruv Kudale is a researcher at the Indian Institute of Technology Bombay (ORCID 0000-0002-3862-819X) whose reviewed work concentrates on tables inside document images — detecting them, decoding their structure, and answering questions over them. The Evolve catalog holds two 2026 arXiv preprints. The first, EMBLEM: Enhancing Multi-script Table Detection through Masking, is a first-author, corresponding-author paper proposing masking-based fine-tuning for multi-script table detection and introducing the MANDALA benchmark; Evolve reads the reported 20.8 F1 lift as a real gain over pretrained models, though smaller than matched fine-tuning achieves. The second, Tables Decoded: DELTA for Structure, TARQA for Understanding, is a middle-author contribution describing a modular TATR+SPRINT+OCR pipeline that emits OTSL, plus an OTSL-tuned LLM (TARQA) that delivers strong TEDS-S and TabVQA gains even as full TEDS (~55) trails SOTA because of OCR errors. The bibliometric radar scores reproducibility at 100 across both reviewed papers, collaboration at 36 with 5 distinct co-authors, and focus at 0 owing to an even 0.5/0.5 field split between document intelligence and computer-vision document image analysis — two labels that, here, describe the same table-centric theme. Output, impact and activity metrics are not yet available. Both preprints carry Evolve's highest reproducibility-readiness rating, so the profile reads as an early-stage but coherent research agenda with precise, re-testable claims.
Collaboration pattern
Both reviewed papers sit in a small, stable IIT Bombay cluster: 5 distinct co-authors total, with Ganesh Ramakrishnan, Jahanvi Rajput, Saikiran Kasturi, Utkarsh Verma and Udhay Brahmi as recurring names. Kudale leads EMBLEM as first and corresponding author while contributing in a middle position on Tables Decoded — stepped-up leadership inside one group rather than wide external collaboration (radar collaboration score: 36).
Research keywords
Highlights
First-author EMBLEM pushes multi-script table detection
DELTA + TARQA: a pipeline pairing table structure with understanding
Both 2026 preprints rated high reproducibility
Both reviewed papers earn Evolve's 'high' reproducibility readiness, yielding a reproMean of 100 over 2 papers — concrete pipelines (TATR+SPRINT+OCR, masking fine-tuning) and a new benchmark (MANDALA) make the claims re-testable.
Timeline
- 2026Publishes EMBLEM as first and corresponding author, proposing masking-based fine-tuning for multi-script table detection and introducing the MANDALA benchmark (arXiv).
- 2026Co-authors Tables Decoded: DELTA for Structure, TARQA for Understanding — a modular TATR+SPRINT+OCR pipeline with an OTSL-tuned LLM (TARQA) for table question answering (arXiv).
- 2026Both preprints enter the Evolve catalog rated 'high' reproducibility readiness, giving a reproducibility score of 100 and anchoring a table-centric document AI 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.
- Output, impact and activity axes are unscored because the snapshot contains no publication counts, citation metrics, or per-year productivity data.
- The evidence base is limited to two 2026 arXiv preprints reviewed by Evolve; other outputs by this author may exist outside the catalog.
- The reproducibility score of 100 is computed from only 2 reviewed papers (reproPapers: 2), so it may not generalize to the full body of work.
- Country and career-biography details are absent from the input; identity is anchored to ORCID 0000-0002-3862-819X from public bibliographic data.