Goreti Marreiros

Research profile

Research summary

Goreti Marreiros is a computer scientist at Universidade Técnica do Porto (Portugal), profiled from public bibliographic data (OpenAlex) anchored to ORCID 0000-0003-4417-8401. The record spans two decades, with topic clusters led by multi-agent systems and negotiation (69), multi-criteria decision making (51), decision support applications (47), context-aware activity recognition (42), and recommender systems (33). Activity is exceptional — the basis of a perfect 100 score: yearly counts climb from 4 works in 2005 to 32 in 2022 and 37 in both 2025 and 2026, with 85 works in the most recent three-year window. The snapshot reports h-index 21 and i10-index 59, though the output and impact radar axes remain unscored pending per-paper data. Recent work pivots to streaming machine learning: Evolve has reviewed two 2026 arXiv papers, both with Marreiros as last author — a collaborative streaming anomaly-detection system with interactive explanations and ensemble consensus (low reproducibility readiness; one loom, three episodes, no baselines) and streaming hierarchical inference with tabular foundation models (medium; consistent error/communication trade-offs on six streams). The reviewed slice is built with a compact team of 5 distinct co-authors, while a focus score of 0 reflects deliberately broad interests across agents, decisions, and data streams.

Research keywords

Highlights

• Hierarchical TabPFN inference for data streams

  Evolve-reviewed 2026 arXiv paper on hierarchical kNN-gated TabPFN offloading for edge-cloud collaboration; the verdict notes consistent error/communication trade-offs across six streams, with untested statistics and latency claims still open. Reproducibility readiness is rated medium and Marreiros is the final author.

• Interactive ensemble consensus for streaming anomalies

  A 2026 arXiv system combining four anomaly detectors, rule-based surrogate explanations, and human refinement; Evolve's verdict credits solid engineering but finds the evidence thin — one loom, three episodes, and no baselines. Reproducibility readiness is rated low and Marreiros is the final author.

• Two decades across agents, decisions, and streams

  The topic record is led by multi-agent systems and negotiation (69), multi-criteria decision making (51), decision support applications (47), context-aware activity recognition (42), and recommender systems (33), alongside snapshot h-index 21 and i10-index 59 — breadth that explains the focus score of 0.

Timeline

1. 2005Earliest year in the publication record, with 4 works counted.
2. 2015Yearly output steps up to 18 works, the first clear rise in the record.
3. 202232 works recorded, the highest annual count to that point.
4. 202537 works, tied for the highest annual count in the record.
5. 202637 works counted; two arXiv streaming-ML papers enter the Evolve catalog with Marreiros as last author.

Papers in the Evolve catalog

• 2026-09-20

  Collaborative Streaming Anomaly Detection with Interactive Explanations and Ensemble Consensus

• 2026-09-07

  Streaming Hierarchical Inference with Tabular Foundation Models

Frequent co-authors