First author of a 2026 diagnostic isolating perception vs. reasoning failures in multimodal LLMs
1
papers here
1
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-21) 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 axis measures publication volume and cumulative output. The scoped record contains only one 2026 first-author arXiv preprint, so no reliable output score can be computed yet.
Impact—
Not yet scored: this axis tracks citation-based influence, such as citation counts and h-index. The single 2026 preprint is too recent for meaningful citation accrual, so no impact estimate is made.
Focus100/100
Scores 100/100 because every catalogued paper maps to a single field: artificial intelligence (multimodal LLM evaluation), a 1-of-1 field distribution. The perfect concentration reflects a deliberately targeted, one-paper record from 2026 rather than a long-honed specialty; as more work is catalogued, this axis will show whether the multimodal-evaluation focus holds or diversifies.
Collaboration36/100
Scores 36/100, computed from 5 distinct coauthors to date. All five — Sree Krishna Uppalapati, Dhruvkumar Patel, Ria Khatoniar, Tanuja Ganu and Rajiv Ratn Shah — connect to the single 2026 first-author preprint, indicating one stable small team rather than a broad network; a larger record is needed before collaborative reach can be fairly judged.
Activity—
Not yet scored: this axis measures publishing cadence and research activity over time. With a single paper in 2026, there is no multi-year series from which an activity trajectory could be derived.
Reproducibility55/100
Scores 55/100, the mean reproducibility-readiness over 1 reviewed paper (reproPapers = 1). The 2026 preprint is rated medium: Evolve's review finds the five-task diagnostic design sound, but flags the absence of significance tests and a confounded geometry-QA (GQ) ceiling — two gaps that directly limit independent replication of the reported results.
Research summary
This is an early-stage researcher profile for Raj Jaiswal, name-resolved from public bibliographic data; institutional affiliation and country are not part of the current record. The scoped evidence base is a single first-author paper in the Evolve catalog: 'Lost in Perception: Isolating Perceptual and Reasoning Failures in Multimodal Physics and Geometry Reasoning', a 2026 arXiv preprint in artificial intelligence (multimodal LLM evaluation). The paper contributes a five-task diagnostic over multimodal physics and geometry question answering that separates perceptual failures from reasoning failures, supported by a chain-of-thought error taxonomy and LLM-as-a-judge scoring. Evolve's review judges the design sound while flagging the absence of significance tests and a confounded GQ ceiling; reproducibility readiness is rated medium, at 55/100. Bibliometric signals are correspondingly narrow: all catalogued output falls in one field (focus 100/100) and the record shows 5 distinct coauthors (collaboration 36/100). The recurring collaborator group — Sree Krishna Uppalapati, Dhruvkumar Patel, Ria Khatoniar, Tanuja Ganu and Rajiv Ratn Shah — suggests a small, stable team. Output, impact and activity lack sufficient data and are reported as missing rather than estimated.
Collaboration pattern
The record shows 5 distinct coauthors, all connected to the single 2026 first-author preprint, giving a collaboration score of 36/100. The recurring group — Sree Krishna Uppalapati, Dhruvkumar Patel, Ria Khatoniar, Tanuja Ganu and Rajiv Ratn Shah — points to one stable small team rather than a wide network; Raj Jaiswal led the work as first author (non-corresponding). With only one catalogued paper, co-authorship roles beyond this project cannot yet be characterized.
Research keywords
Highlights
A five-task diagnostic that separates perception from reasoning
Chain-of-thought error taxonomy with LLM-as-a-judge
Medium reproducibility with clearly flagged gaps
Evolve rates the paper's reproducibility readiness as medium (55/100): the review credits a sound design but notes missing significance tests and a confounded GQ ceiling that future revisions should address.
Timeline
- 2026Posts first-author arXiv preprint 'Lost in Perception: Isolating Perceptual and Reasoning Failures in Multimodal Physics and Geometry Reasoning'.
- 2026The paper enters the Evolve catalog, reviewed as a sound five-task diagnostic with flagged gaps (no significance tests; confounded GQ ceiling; reproducibility medium, 55/100).
- 2026Publishes with a five-person collaborator group on the single-paper record (5 distinct coauthors; collaboration score 36/100).
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 dimension scores are computed from public bibliometric data per Evolve's radar methodology; venue and affiliation metadata may be incomplete.
- Profile scope covers a single reviewed paper (2026 arXiv preprint); output, impact and activity are reported as missing, not zero.
- Institution and country are unavailable; 'Raj Jaiswal' is name-resolved from public bibliographic data via ORCID 0009-0008-5044-1409, and identity with other same-name researchers cannot be ruled out.
- The reproducibility score reflects one Evolve review of one paper and may change as more work is catalogued.