Jeremy C. Weiss

National Institutes of Health · United States

Unclaimed · AI-structured profileOpenAlex

NIH clinical NLP researcher: clinical timeline extraction, temporal reasoning, and grounded LLM-as-a-judge evaluation.

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-15) 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—

Flagged as missing: this OpenAlex identity reports an h-index and i10-index of 0 with only a 2026 count (12 works), so lifetime publication volume cannot be measured. This axis would track cumulative output and its trajectory.

Impact—

Impact is also flagged missing: no usable citation figures exist in the snapshot (h-index and i10-index recorded as 0). This axis would measure citations, scholarly uptake, and the work's influence on clinical NLP practice.

Focus0/100

The radar scores focus at 0, the lowest axis in this snapshot. The raw field split is actually narrow — three near-identical clinical NLP labels at one third each — but the OpenAlex topic profile spans five areas (machine learning in healthcare 12, topic modeling 7, biomedical text mining and ontologies 6, EHR systems 2, AI in healthcare and education 2), which the scoring reads as dispersion across themes.

Collaboration39/100

Collaboration scores 39, the weakest measured axis. The snapshot counts 6 distinct coauthors, matching the six recurring names: Sayantan Kumar, Jack Cummins, Nicolas Grimaldi, Misaki Matsuura, Ojas Kadam, and Ketan Tamirisa. All three catalog papers place Weiss in the last author position, though none as corresponding author — senior involvement within a small, stable writing group rather than a broad network.

Activity75/100

Activity is the strongest measured axis at 75. OpenAlex records 12 works in 2026, and the same 12 constitute the three-year window (recent3yWorks 12) — recent output is concentrated entirely in this single year, including the three cataloged arXiv papers. The picture is a burst of current productivity whose longer-term cadence cannot yet be verified from this record.

Reproducibility55/100

Reproducibility sits mid-range at 55, derived from Evolve's review of 3 papers whose reproducibility readiness is each rated medium. The reviews point to concrete gaps rather than fundamental flaws: GAVEL's merge gains reuse its own judging instrument with judge recall unmeasured, and the multimodal reconstruction study reports n=40 with a 40%-vs-43% recovery-claim mismatch. Artifacts look plausible but remain under-validated.

Research summary

Jeremy C. Weiss is a researcher at the US National Institutes of Health working in clinical natural language processing and biomedical informatics. His cataloged 2026 output — three arXiv papers, each with him in the last (senior) author position — targets one coherent problem: making clinical time machine-readable and machine-judgeable. GAVEL introduces a case-report-grounded LLM judge for comparing extracted clinical timelines; a paired benchmark shows that exposing full timelines shifts LLM clinical answers toward hindsight traps without accuracy gains; and a UID-preserving multimodal reconstruction method anchors discharge-summary events in time with source-grounded adjudication. OpenAlex records 12 works in 2026, all within the recent three-year window, against a compact pool of 6 distinct coauthors. Evolve rates all three papers' reproducibility readiness as medium (mean 55), flagging judging-instrument reuse, GPT-generated labels, and small-n ablations as limitations. The snapshot's citation fields (h-index and i10-index recorded as 0, countsByYear limited to 2026) indicate an incomplete identity record rather than an established impact profile, so output and impact remain unassessed. This is an active, tightly focused early-stage profile in clinical temporal NLP and LLM evaluation.

Collaboration pattern

Weiss writes within a compact, stable group: the snapshot counts 6 distinct coauthors — Sayantan Kumar, Jack Cummins, Nicolas Grimaldi, Misaki Matsuura, Ojas Kadam, and Ketan Tamirisa — and he holds the last author position on all three 2026 catalog papers, none as corresponding author. The collaboration score of 39 reflects this small recurring cluster rather than a wide network; the papers' shared GAVEL and evaluation infrastructure suggests one core team iterating across studies.

Research keywords

Clinical timeline extractionTemporal reasoningClinical temporal reasoningLLM-as-a-judgeGrounded evaluationHindsight biasCase reportsDischarge summariesMultimodal EHRRetrieval-augmented generationBiomedical text miningBenchmark evaluation

Highlights

GAVEL: case-report-grounded LLM judging of clinical timelines
Introduces a grounded adjudication framework that compares extracted clinical timelines against their source case reports, with Bradley-Terry-based comparison. Evolve's review calls it well validated, but notes that the merge gains reuse the judging instrument itself and that judge recall is unmeasured.
Paired benchmark exposes hindsight bias in LLM clinical judgment
Shows that giving models the full clinical timeline shifts their answers toward hindsight traps without an accuracy gain, using temporal masking over textual time series. Evolve cautions that GPT-generated labels and LLM judging temper the evidence.
UID-preserving multimodal timeline reconstruction anchored by GAVEL
Couples discharge summaries with retrieval-augmented multimodal reconstruction to keep event identities stable across the timeline, then adjudicates with the GAVEL judge. Evolve credits rigorous ablations but flags n=40 and a 40%-vs-43% recovery-claim mismatch.

Timeline

2026Publishes UID-preserving multimodal clinical timeline reconstruction with source-grounded GAVEL adjudication.
2026Releases a paired benchmark showing hindsight bias in LLM clinical temporal reasoning under full-timeline exposure.
2026Introduces GAVEL, a report-grounded LLM judge for comparing extracted clinical timelines.
2026OpenAlex logs 12 works in the year, the entirety of the recorded three-year output.

Papers in the Evolve catalog

Frequent co-authors

Sayantan Kumar×3 Jack Cummins×2 Nicolas Grimaldi×1 Misaki Matsuura×1 Ojas Kadam×1 Ketan Tamirisa×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 bibliometrics come from an OpenAlex snapshot; no ORCID was available, so this profile is name-resolved from public bibliographic data and could merge or omit namesakes.
Citation fields are effectively empty (h-index and i10-index recorded as 0; countsByYear covers only 2026), so lifetime output and impact could not be assessed.
The three catalog papers are 2026 arXiv preprints; peer-review status and venue acceptance are not captured in the input data.
The reproducibility score (55) reflects Evolve's review of 3 papers rated medium readiness, not an independent audit of released code or artifacts.