An Explainable Agentic System for Detection of Conversational Scams with Summary-Based Memory
3.00T1 sourcearXiv cs.MA
Source record
Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2607.11707.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryPaper presents an explainable agentic system with summary-based memory for detecting multi-turn conversational scams across eight types, releases the ConScamBench-278 benchmark, reports 100% phishing recall on isolated messages, 97.8% accuracy on the benchmark, and positive results from two user studies (N=100, N=45) with SUS score 74.7.
Why it mattersThe summary-based memory design and a public multi-category scam benchmark are concrete artifacts other builders of agentic detection pipelines can reuse or compare against.
Cited by
No citations on record.