Blockchain forensics is adversarial and open-ended, according to a new arXiv paper introducing ABF-Bench. Money launderers split, swap, and bridge funds, cut their trail at mixers or custodial services, and change tactics from one incident to the next. Investigators, whether human or automated, must keep adapting.
ABF-Bench is designed as a verifiable benchmark for agentic blockchain forensics, meaning the correctness of an agent's investigation can be checked rather than judged by loose qualitative criteria. The paper argues that this kind of evaluation is needed because forensic work is not a closed problem with a single fixed answer.
Since only one source was provided, there are no differing accounts to reconcile. The abstract's main claim is that ABF-Bench offers a way to test AI agents under realistic, adversarial conditions.