Document intelligence researcher at IIT Bombay leading work on table structure recognition and table question answering
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: this profile draws only on the two 2026 arXiv papers reviewed by Evolve, without full publication counts. This axis would measure overall research output — paper volume, venues, and authorship roles across the complete record.
Impact—
Not yet scored: no citation counts, h-index, or venue-level influence data was provided. This axis would measure uptake of the work — citations, and downstream use of resources such as the OTSL pipeline and the MANDALA benchmark.
Focus0/100
The radar returns a focus score of 0. The underlying field distribution is split evenly (0.5 / 0.5) between document intelligence — computer vision and NLP for table understanding — and computer vision for document image analysis. In practice the two reviewed papers address adjacent table-centric problems (structure plus QA, and multi-script detection), so the low score reflects a two-pronged table agenda rather than scattered interests.
Collaboration36/100
The collaboration score is 36, computed from 5 distinct coauthors in the current snapshot. All five — Dhruv Kudale, Jahanvi Rajput, Saikiran Kasturi, Utkarsh Verma, and Udhay Brahmi — recur across the reviewed work, and on both 2026 arXiv papers Ramakrishnan sits in the last author position (non-corresponding), a pattern consistent with a senior supervisor guiding a compact, student-led team rather than a large distributed consortium.
Activity—
Not yet scored: the visible record spans a single year, 2026, so publishing cadence cannot be established. This axis would measure ongoing activity — publication frequency, recency, and continuity across multiple years.
Reproducibility100/100
Reproducibility is the strongest axis, scoring 100 (reproMean 100 across the 2 reviewed papers). Evolve rates both the DELTA/TARQA table pipeline and EMBLEM's masking approach as high reproducibility readiness: the modular TATR+SPRINT+OCR design and the MANDALA benchmark with reported F1 numbers give third parties concrete artifacts to rebuild and stress-test the reported gains.
Research summary
Ganesh Ramakrishnan is a researcher at the Indian Institute of Technology Bombay whose recent work centers on document intelligence — the vision and NLP machinery that makes tables in documents machine-readable. The Evolve catalog holds two 2026 arXiv papers, both with him in the last author position. The first, 'Tables Decoded: DELTA for Structure, TARQA for Understanding,' couples a modular TATR+SPRINT+OCR pipeline (DELTA) that emits an OTSL representation with an OTSL-tuned LLM (TARQA); the pairing delivers strong TEDS-S and TabVQA gains, although full-table TEDS of roughly 55 still trails the state of the art because of OCR errors. The second, 'EMBLEM: Enhancing Multi-script Table Detection through Masking,' reports that masking-based fine-tuning lifts multi-script table detection by 20.8 F1 on the new MANDALA benchmark; Evolve's review judges this a real gain over pretrained baselines, though smaller than matched fine-tuning. Bibliometrically, the snapshot is compact: 5 distinct coauthors (Dhruv Kudale, Jahanvi Rajput, Saikiran Kasturi, Utkarsh Verma, Udhay Brahmi), a field split of 0.5/0.5 between document intelligence and document image analysis, and a perfect reproducibility score of 100 across the two reviewed papers. Output, impact, and activity axes remain unscored for lack of data. The profile is name-resolved from public bibliographic data under ORCID 0000-0003-4533-2490.
Collaboration pattern
Within this snapshot Ramakrishnan works with a compact circle: 5 distinct coauthors, all recurring — Dhruv Kudale, Jahanvi Rajput, Saikiran Kasturi, Utkarsh Verma and Udhay Brahmi. On both 2026 arXiv papers he takes the last author position (non-corresponding), consistent with a senior, supervisory role over a student-led team. The collaboration score of 36 reflects this small, stable core rather than a broad international network; all visible output sits at IIT Bombay.
Research keywords
Highlights
Modular DELTA pipeline turns tables into OTSL for LLM consumption
EMBLEM: masking boosts multi-script table detection on new MANDALA benchmark
High reproducibility readiness across both reviewed papers
Both 2026 papers carry a 'high' reproducibility readiness rating from Evolve, yielding a reproducibility score of 100 (reproMean 100 over 2 papers) — modular pipeline design and a new public benchmark make the results straightforward to rebuild and test.
Timeline
- 2026Last-author arXiv paper 'Tables Decoded: DELTA for Structure, TARQA for Understanding' presents the DELTA pipeline and TARQA model for table structure recognition and question answering.
- 2026Last-author arXiv paper 'EMBLEM: Enhancing Multi-script Table Detection through Masking' introduces masking-based fine-tuning and the MANDALA multi-script benchmark.
- 2026Both papers enter the Evolve catalog with 'high' reproducibility readiness, giving a reproducibility score of 100 across the 2 reviewed papers.
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.
- Bibliometric scores derive from a very small snapshot: 2 Evolve-reviewed 2026 arXiv papers and 5 distinct coauthors; they may not represent the researcher's full publication record.
- Output, impact and activity dimensions are unscored — no citation counts, h-index, venue history or multi-year publication data was provided.
- The reproducibility score of 100 covers only the 2 reviewed papers (reproMean 100), not the wider portfolio.
- Profile is name-resolved from public bibliographic data and matched to ORCID 0000-0003-4533-2490; common-name ambiguity beyond that scope cannot be excluded.