Hanoz Bhathena

JPMorgan Chase & Co (United States)

Unclaimed · AI-structured profileOpenAlex

JPMorgan Chase researcher advancing LLM-based evaluation and coverage auditing for production RAG retrieval

papers here

reviewed by Evolve

first-author

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 metrics are not yet available: no publication counts, venue history, or yearly production totals underlie this axis. Once the bibliographic record is linked to a persistent identifier, this axis would measure the volume and consistency of scholarly output.

Impact—

Impact data is missing: no citation counts, h-index, or downstream-usage records are attached to this profile. This axis would measure how widely the two cataloged RAG-evaluation papers are cited, reused, or adopted in production systems.

Focus0/100

Focus scores 0 because the field distribution splits evenly (0.5/0.5) across two labels: natural language processing / information retrieval, and information retrieval (LLM-based evaluation for RAG). In substance the halves are adjacent — both papers target evaluation of retrieval for RAG — so the topical spread is narrower than the 0 score implies. The score is largely an artifact of field-label granularity on a two-paper sample.

Collaboration32/100

The collaboration dimension scores 32, computed from a small co-authorship sample: four distinct co-authors appear across the two cataloged papers, with Aviral Joshi, Saket Sharma, and Max Nelson recurring most often (Max Nelson is listed twice in the raw record). Bhathena sits in a middle author position on both papers, not first or corresponding, indicating contribution within a team rather than independent leadership. The low score likely reflects the two-paper sample size more than collaboration habits.

Activity—

Activity data is insufficient: the cataloged papers all carry 2026 timestamps, so no multi-year cadence can be computed. This axis would measure publication frequency, latency between outputs, and research continuity across years.

Reproducibility55/100

Reproducibility scores 55, the mean readiness rating across the two scored papers, each individually rated medium. Evolve's reviews support the middle band: Re:CAP is described as unusually rigorous and honest despite a TREC-COVID recall regression, while the pooled-evaluation paper shows solid validation limited by sparse gold qrels. Neither paper reaches a high readiness flag, so 55 reads as moderately reproducible, resting on only two data points.

Research summary

Hanoz Bhathena is a researcher at JPMorgan Chase & Co (United States) working at the intersection of natural language processing and information retrieval, with a concentrated interest in evaluation methodology for retrieval-augmented generation (RAG). The Evolve catalog holds two 2026 arXiv papers, on both of which Bhathena is a middle-listed author and not the corresponding author. The first, Re:CAP, audits retrieval coverage in production RAG pipelines using reference-free LLM probing; Evolve's verdict calls it unusually rigorous and honest, while noting that TREC-COVID recall regresses. The second proposes incremental pooled LLM evaluation for retrieval model selection, claiming stable selection at 4.9x lower cost; validation is described as solid, but sparse gold qrels limit certainty. Bibliometric signals rest on this small sample: four distinct co-authors and a collaboration score of 32; a field split of 0.5/0.5 between NLP/IR and LLM-based retrieval evaluation, which produces a focus score of 0 even though both halves are adjacent subfields; and a reproducibility mean of 55 across two papers rated medium. Output, impact, and activity metrics are not yet available. The profile is name-resolved from public bibliographic data; 'Hanoz Bhathena' may not be unique.

Collaboration pattern

The record shows a compact collaboration footprint: four distinct co-authors across two papers, with Aviral Joshi, Saket Sharma, and Max Nelson recurring most frequently (Max Nelson appears twice in the raw list). Both papers carry the JPMorgan Chase industry affiliation, with Bhathena in a middle author position rather than first or corresponding. The collaboration score of 32 likely reflects this small, concentrated, team-internal sample rather than an extensive external network.

Research keywords

retrieval-augmented generation (RAG)retrieval evaluationLLM-as-judgecoverage auditingiterative query probingproduction monitoringtest collection poolingtest collection reuseretrieval model selectionevaluation cost efficiencyproduction RAG pipelinesreference-free probing

Highlights

Re:CAP: auditing retrieval coverage in production RAG pipelines
The 2026 arXiv paper claims that reference-free LLM probing recovers gold documents that flat RAG retrieval misses. Evolve's verdict calls the work unusually rigorous and honest, while noting that recall on TREC-COVID regresses; reproducibility readiness is rated medium and Bhathena is a middle author.
4.9x cheaper retrieval model selection via incremental pooled LLM judging
The second 2026 arXiv paper claims that pooled LLM judging with incremental pool growth yields stable retrieval model selection at 4.9x lower cost. Validation is described as solid, though sparse gold qrels limit certainty; readiness is medium and Bhathena is a middle author.
A coherent LLM-as-judge evaluation agenda for RAG retrieval
Across both cataloged 2026 papers, one thread recurs: using LLM judges to evaluate retrieval for RAG — from live coverage monitoring in production on one side and cost-efficient model selection on the other — within an industry setting at JPMorgan Chase.

Timeline

2026Published Re:CAP on arXiv — an audit framework for retrieval coverage in production RAG pipelines, in a middle author position.
2026Published incremental pooled LLM evaluation for cost-effective retrieval model selection on arXiv, also in a middle author position.
2026Both cataloged papers land in the same year, establishing an early but concentrated agenda around LLM-based evaluation of RAG retrieval.

Papers in the Evolve catalog

Frequent co-authors

Aviral Joshi×2 Saket Sharma×2 Max Nelson×1 Max Nelson×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.

All bibliometric dimensions derive from a very small sample: only two cataloged 2026 arXiv papers; scores should not be read as career-level statistics.
Output, impact, and activity axes have no underlying data (no citation counts, publication totals, or multi-year history), and no ORCID is linked to this profile.
The profile is name-resolved from public bibliographic data; 'Hanoz Bhathena' may not be a unique name, and the affiliation reflects the snapshot record.
The field distribution uses only two coarse labels (0.5/0.5), so the focus score of 0 reflects label granularity on a two-paper sample rather than demonstrated breadth.