Overview

Retrieval

Modification

InspectAI Evaluation Logs

Every number on this dashboard traces to a raw InspectAI evaluation log. Browse the full cross-machine corpus in the hosted viewer.

View all InspectAI evaluation logs →

Hosted on a HuggingFace Docker Space. The viewer may take ~30 s to wake from cold start on first visit.

About

Citation

aLLarMa: A Two-Stage Constrained GraphRAG Framework for Power System Alarm, Network and Operational Data Analytics

Authors: Saumil Shah (NC State University; Electric Power Research Institute), Adrian Kelly (Electric Power Research Institute, Ireland), Wenyuan Tang (NC State University)

Hardware

  • SKORGE (consumer-grade workstation): AMD Ryzen 9 7950X CPU, NVIDIA RTX 4090 GPU (24 GB), 128 GB RAM, vLLM serving.
  • NVIDIA DGX Spark (self-contained 240 W edge-class appliance): GB10 Grace–Blackwell superchip (20-core ARM CPU + 5th-gen Tensor-Core GPU, 128 GB unified memory), vLLM serving.