Tajana Rosing
University of California San Diego · United States
UC San Diego researcher in LLM agents and time-series QA; high recent activity, strong reproducibility
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
1
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 data is insufficient for scoring: cumulative publication volume, venue mix, and full-record yearly counts are not yet indexed for this identity. This axis would measure total and per-year paper output across the complete publication record.
Impact—
Insufficient data for scoring: citation counts and downstream-usage signals are unavailable; the snapshot records only h-index 1 and i10-index of 0. This axis would measure scholarly influence through citations and adoption of the researcher's publications.
Focus0/100
Focus scores 0/100, the weakest axis: the field distribution splits evenly at 0.5 between machine learning (LLM agents) and LLM-based time-series question answering, and the record spans five OpenAlex topic clusters — Topic Modeling (7), Parallel Computing and Optimization Techniques (6), Advanced Database Systems and Queries (4), Advanced Memory and Neural Computing (4), and Domain Adaptation and Few-Shot Learning (3). Activity is real, but thematically dispersed.
Collaboration44/100
Collaboration scores 44/100 on a moderate base: 8 distinct coauthors appear across recent work, with a recurring cluster including Onat Gungor, Siddharth Sharma, and Nilesh Prasad Pandey. Both Evolve-reviewed 2026 papers place this researcher in the last (senior) author position, consistent with a supervisory role, though neither lists them as corresponding author.
Activity89/100
Activity scores 89/100, the strongest axis: 20 works in the recent three-year window, with OpenAlex countsByYear rising from 4 works in 2025 to 16 in 2026 — a fourfold jump. Both catalog papers appeared on arXiv in 2026, indicating an accelerating, currently intensive publication cadence rather than a historical backlog.
Reproducibility78/100
Reproducibility scores 78/100, computed from Evolve's review of 2 papers (mean 78). T-SMART is rated high reproducibility readiness, backed by rigorous paired statistical comparisons and explicit ablation studies; LIMBO is rated medium, its LinUCB-based allocator described as lightly ablated. Overall, the record favors verifiable, mechanism-level evaluation.
Research summary
Tajana Rosing is a researcher at the University of California San Diego whose indexed record concentrates on machine learning, split evenly (0.5 each in the field distribution) between LLM-based agents and LLM-based time-series question answering. The OpenAlex snapshot is compact and recent: 20 works in the last three-year window, with countsByYear rising from 4 in 2025 to 16 in 2026 — an activity score of 89/100, the profile's strongest axis. Citation metrics are thin — h-index 1 and i10-index 0 — indicating the indexed record spans only 2025–2026 and is not yet a complete career record. Evolve has reviewed two 2026 arXiv papers with Rosing in the last (senior) author position, neither as corresponding author. T-SMART proposes mechanism-level attribution for tool-augmented time-series question answering, crediting performance gains to deterministic tools rather than LLM reasoning, with rigorous paired statistics and high reproducibility readiness; Evolve notes the results are taxonomy-scoped and trail TS-Agent. LIMBO introduces lifelong inference-time memory and budget optimization for LLM agents through a LinUCB contextual bandit, matching fixed-replay accuracy at 26–83% lower cost, though lightly ablated. The reproducibility axis scores 78/100 across these two papers. Focus scores 0/100: effort spreads across five topic clusters, led by Topic Modeling (7) and Parallel Computing and Optimization Techniques (6). Collaboration is moderate (44/100) with 8 distinct coauthors, including frequent collaborators Onat Gungor and Siddharth Sharma.
Collaboration pattern
Collaboration is moderate (44/100) with 8 distinct coauthors in the recent window. Both Evolve-reviewed 2026 papers place Rosing in the last author position — consistent with a senior or supervisory role — though not as corresponding author. A recurring coauthor cluster includes Onat Gungor, Siddharth Sharma, and Nilesh Prasad Pandey, alongside Ivan Delgado, Himansi Gupta, Bishal Khatri, Niharika Sapre, and Lameta Shamoon.
Research keywords
Highlights
T-SMART: gains in tool-augmented time-series QA traced to deterministic tools
LIMBO: bandit-tuned memory and budget cut LLM-agent inference cost 26–83%
Timeline
- 2025First fully indexed year in the OpenAlex snapshot: 4 works recorded under this identity.
- 2026Publication count jumps fourfold to 16 works, the peak year in the snapshot.
- 2026T-SMART appears on arXiv, introducing mechanism-level attribution for tool-augmented time-series question answering.
- 2026LIMBO appears on arXiv, presenting lifelong inference-time memory and budget optimization for LLM agents.
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 figures come from a single OpenAlex identity; the indexed record spans 2025–2026 (20 works), so citation metrics (h-index 1, i10-index 0) reflect a partial record rather than a complete career.
- No ORCID was provided; the profile is name-resolved from public bibliographic data.
- Output and impact dimensions lack sufficient underlying data and are left unscored.
- Reproducibility (mean 78) is computed from Evolve's review of only 2 papers.