LLM agents and time-series QA researcher at West Virginia University with a sharp 2025–2026 output ramp
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
8
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—
Output metrics are not yet computed for this profile, so no score is shown. This axis would measure publication volume and productivity — annual output counts and venue distribution — with the underlying data currently insufficient in the snapshot.
Impact—
Impact data is insufficient in this snapshot, so no citation-based score is reported. This axis would measure research influence through citation counts, citations per paper, and citing works; it will be scored once citation records are available.
Focus0/100
The focus score of 0 reflects a highly dispersed topical profile. Historical topics span five areas: adversarial robustness in machine learning (28), network security and intrusion detection (23), topic modeling (10), artificial immune systems applications (10), and security and verification in computing (9). The current field distribution splits evenly (50%/50%) between LLM agents and LLM-based time-series question answering, showing breadth across security and LLM systems rather than a concentrated line.
Collaboration44/100
The collaboration score of 44 sits in the middle range and matches a compact working pattern: 8 distinct co-authors appear across 46 works in the last three years. Recurring partners include Tajana Rosing, Siddharth Sharma, Nilesh Prasad Pandey, Ivan Delgado, Himansi Gupta, Bishal Khatri, Niharika Sapre, and Lameta Shamoon. On both cataloged 2026 papers (T-SMART and LIMBO) he is a middle author and not the corresponding author, indicating contribution within team-led projects rather than solo leadership.
Activity100/100
The activity score of 100 is the strongest signal in the profile. OpenAlex counts by year show 1 work in 2019, 2 in 2020, 6 in 2021, 3 in 2022, 6 in 2023, and 5 in 2024, followed by a sharp ramp to 20 works in 2025 and 21 in 2026; 46 works fall within the last three years. Both Evolve-reviewed papers are 2026 arXiv preprints, consistent with this surge.
Reproducibility78/100
The reproducibility score of 78 is the mean over 2 assessed papers. T-SMART rates high readiness, with rigorous paired statistical testing supporting its mechanism-level attribution claims, though Evolve notes it is taxonomy-scoped and trails TS-Agent. LIMBO rates medium readiness: its LinUCB-based memory and budget allocation matches fixed-replay accuracy at 26–83% lower cost, but the paper is lightly ablated. Together they indicate solid statistical practice with room for fuller ablation coverage.
Research summary
Onat Gungor is a researcher at West Virginia University (US) whose recent work centers on large language model agents and LLM-based time-series question answering, each accounting for 50% of his current field distribution in the Evolve snapshot. His bibliometric record (OpenAlex) shows an h-index of 8, an i10-index of 7, and 46 works over the last three years, with annual output climbing from 5 works in 2024 to 20 in 2025 and 21 in 2026 — a steep ramp that drives his activity score of 100. Earlier topical data point to a security-flavored background: adversarial robustness in machine learning (28 records) and network security and intrusion detection (23) dominate his topic profile, alongside topic modeling, artificial immune systems applications, and security and verification in computing. The Evolve catalog reviews two 2026 arXiv papers: T-SMART, which attributes tool-augmented time-series QA gains to deterministic tools rather than LLM reasoning, and LIMBO, which uses a LinUCB contextual bandit to cut inference cost by 26–83% at matched accuracy. Both list him as a middle author, reflecting contribution within a compact pool of 8 distinct recent co-authors. His reproducibility score of 78 (mean over 2 assessed papers) is anchored by T-SMART's rigorous paired statistics and high reproducibility readiness.
Collaboration pattern
With 8 distinct co-authors across 46 works in the last three years, Gungor works in a compact, stable team pattern (collaboration score 44). His most frequent collaborators include Tajana Rosing, Siddharth Sharma, Nilesh Prasad Pandey, Ivan Delgado, Himansi Gupta, Bishal Khatri, Niharika Sapre, and Lameta Shamoon. On both cataloged 2026 papers he appears as a middle, non-corresponding author, suggesting he contributes specialist components within senior-led groups rather than running his own lab's output.
Research keywords
Highlights
T-SMART pins tool-augmented time-series QA gains on deterministic tools
LIMBO cuts LLM-agent inference cost 26–83% with LinUCB allocation
Output ramps from 5 works (2024) to 20 (2025) and 21 (2026)
OpenAlex counts by year show the steepest growth in the record: 20 works in 2025 and 21 in 2026 after 5 in 2024, with 46 works in the last three years and an activity score of 100. Both Evolve-reviewed papers are 2026 arXiv preprints from this surge.
Timeline
- 2019First work recorded in the OpenAlex snapshot (1 work).
- 2021Output reaches 6 recorded works, the highest of the 2019–2024 phase.
- 20245 recorded works — the last year before the output ramp begins.
- 2025Output jumps to 20 recorded works, the steepest single-year increase on record.
- 202621 recorded works; T-SMART and LIMBO, two arXiv preprints on LLM agents and time-series question answering, enter the Evolve catalog.
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 indicators (h-index, i10-index, counts by year, topics) come from an OpenAlex identity snapshot only.
- Output and impact dimension scores are missing from the radar; citation-level data was insufficient and no values were estimated.
- Qualitative assessments rest on only 2 Evolve-reviewed papers (both 2026 arXiv preprints with middle authorship), so depth is limited.
- Profile is name-resolved from public bibliographic data (ORCID 0000-0001-7215-0890); biography, awards, and roles beyond the data are not covered.