Get Started
Home
Topics
Search
Library
Research questionHow can transaction-graph screening detect risky blockchain addresses on chains without labels while keeping alert volume controlled?Sanctions registries leave many blockchain addresses unlabelled and provide no coverage for some chains. Screening must infer risk from transaction relationships while limiting investigator workload and identifying activity before public designation.
AI
Finance
Machine Learning
Research Paper
Statistical Machine Learning
Technology
Latest papersRecent research connected to this question, newest first.Population-Calibrated Graph Screening at 835-Million-Address Scale, with Label-Free Transfer to New ChainsThe evidence describes a deployed graph-screening system over 835 million addresses and 15.8 billion edges spanning five EVM chains. It uses a shared inductive encoder with per-chain normalization, transfers heads trained on two chains to unlabeled target chains, sets thresholds by exact population quantiles, and reports held-out recall, lead-time replay, bit-identical serving at 151 ms p50, drift monitoring, and a measured blind spot against synthesised adversarial behaviour.research paper · Sep 2, 2026
Related questions
How can urban trajectory systems screen collective anomalies cheaply yet provide source-verifiable event details on demand?How can detect-to-track pipelines adapt to new visual domains without target-domain labels?Can chain-of-thought monitoring detect consequential computation hidden in semantically irrelevant filler tokens?How can smaller models reliably detect risk in obfuscated Chinese SMS under low-latency constraints?