Reproduce “AMLBench: A Real-Incident Benchmark for Agentic Blockchain Anti-Money Laundering” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Qishuang Fu, Hang Zheng, Yuchen Lei, Yuexin Xiang, Qin Wang, Yuanzhe Zhang, Shiping Chen, Andreas Deppeler and 3 more
AMLBench tests whether agents can trace incident-related Ethereum transfers from limited clues while recording their investigation. Agents query records, inspect candidate fund-flow links, choose branches, and submit relevant transfers across 24 cases containing 5.05 million transfers. GPT-6 Astra reaches 80.82% macro-average transfer precision and 84.73% query coverage, but large cases remain incomplete: Bybit recall is 0.98%.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.