NSWC Panama City researcher spanning SAS sonar target recognition and transformer interpretability for LLM safety
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-14) 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—
OpenAlex returns no usable output totals: the recorded h-index and i10-index are zero and no aggregate output score is computed. This axis would measure cumulative publication volume and productivity over time; current data is insufficient to assess it.
Impact—
No citation-impact data is available in this snapshot; the zero h-index likely reflects an incomplete OpenAlex profile rather than measured influence. This axis would track citations, citing works, and field-normalized attention once coverage improves.
Focus0/100
The focus score is 0 because OpenAlex's field distribution is an exact 50/50 split: half the record sits in computer vision for underwater sonar sensing (SAS automatic target recognition) and half in machine learning for LLM interpretability and safety. The two 2026 catalog papers confirm this dual track, and topic counts (2 each for SAR imaging, adversarial robustness, underwater acoustics, and XAI; 1 for topic modeling) show no single dominant theme.
Collaboration41/100
Collaboration scores 41 on a record listing 7 distinct coauthors. The frequent-collaborator set — Gregory D. Vetaw, Dan Zimmerman, Amelia L. McCormack, Matthew Cook, Darshan W. Bryner, Matthew G. Cook, and Jose L. Salas-Vernis — forms a compact network consistent with a government-lab team pattern. Roles alternate: first author on Deep Noir (2026) but a middle authorship on the LoRA SAS paper, indicating both lead and contributing positions on small teams.
Activity40/100
Activity scores 40 from a thin but recent record: OpenAlex counts 3 works in the last three years, all logged in 2026, after a single work in 2023 in counts-by-year. Both 2026 Evolve-reviewed papers are arXiv preprints, so current output is concentrated in one burst. The gap between 2023 and 2026 means a long-term productivity cadence cannot yet be established.
Reproducibility38/100
Reproducibility averages 38 across the 2 Evolve-reviewed papers. Deep Noir rates medium: logit-lens and head-attribution results come with ablations, but key automated baselines and code are missing. The SAS LoRA paper rates low despite rigorous matched-control nulls, because private sonar data blocks replication. The pattern is solid internal methodology with weak external artifacts — code and data release would move this score most.
Research summary
Frank E. Bobe III is a researcher at NSWC Panama City Division whose OpenAlex record splits evenly — 50/50 — between two lines: computer vision for underwater sonar sensing (SAS automatic target recognition) and machine learning for LLM interpretability and safety. In 2026 the profile includes a first-authored arXiv preprint, 'Deep Noir,' which automates activation-steering discovery in transformer models via logit-lens reads and attention-head attribution, and a coauthored study, 'LoRA Enhanced Contrastive Learning with SAS Vision Transformers,' reporting that LoRA alone lifts SAS ATR AUPRC from 0.300 to 0.679 while rigorous matched-control nulls show mining and SupCon add nothing. OpenAlex topic tags reinforce the spread: Advanced SAR Imaging Techniques, Adversarial Robustness in Machine Learning, Underwater Acoustics Research, and Explainable AI each appear twice, with Topic Modeling once. The profile is early-stage: 3 works in the last three years, 7 distinct coauthors, and an h-index recorded as 0, indicating sparse bibliographic coverage rather than an established citation record. Evolve's reproducibility assessment across the two reviewed papers averages 38, with medium readiness for Deep Noir (missing automated baselines and code) and low for the SAS study (private data limits replication).
Collaboration pattern
The record lists 7 distinct coauthors in a compact set: Gregory D. Vetaw, Dan Zimmerman, Amelia L. McCormack, Matthew Cook, Darshan W. Bryner, Matthew G. Cook, and Jose L. Salas-Vernis. Matthew Cook and Matthew G. Cook are likely the same collaborator recorded twice. Authorship is mixed — first author on Deep Noir, middle position on the SAS LoRA paper — consistent with alternating lead and supporting roles on small lab teams.
Research keywords
Highlights
Deep Noir: automated steering discovery via logit lens and head attribution
LoRA alone lifts SAS ATR AUPRC from 0.300 to 0.679
Timeline
- 2023First publication activity appears in OpenAlex counts-by-year, with a single work recorded for 2023.
- 2026Three works are recorded for 2026, the most productive year in the record to date.
- 2026Publishes 'Deep Noir' on arXiv as first author, automating steering discovery in transformer models.
- 2026Coauthors the SAS Vision Transformer LoRA study in a middle authorship position.
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 data is drawn from OpenAlex only; no ORCID or verified identity link exists, so this profile is name-resolved from public bibliographic data.
- h-index and i10-index are recorded as 0 with no citation data, indicating an incomplete OpenAlex profile rather than measured zero impact.
- The reproducibility score (mean 38) rests on only 2 Evolve-reviewed arXiv preprints; no journal publications are in the reviewed catalog.
- The 2023 entry appears as a single unattributed work in counts-by-year; that paper is not in the Evolve catalog, so its content is unknown.