PreScam is a new benchmark showing supervised models outperform zero-shot LLMs at predicting when scam conversations approach termination but struggle with forecasting next scammer actions.
If a round has no valid PT after review, merge it into an adjacent round or move it to initial_contact or outcome
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
PreScam: A Benchmark for Predicting Scam Progression from Early Conversations
PreScam is a new benchmark showing supervised models outperform zero-shot LLMs at predicting when scam conversations approach termination but struggle with forecasting next scammer actions.