New ReDence framework rapidly detects social engineering scams with 96% accuracy

Researchers have developed ReDence, an open-source framework designed to rapidly detect social engineering scams by grounding its decisions in evidence. The system employs a hybrid approach, utilizing a fine-tuned multimodal agentic model called Investigator (based on Qwen3.5-4B) for planning and interpretation, and a specialized Decider model (Tev1) for specific decision-making tasks. This architecture aims to improve detection speed and accuracy, with ReDence achieving 96% accuracy across twelve scam categories and being at least four times faster than comparable agentic models. AI

IMPACT This framework could significantly improve the speed and accuracy of detecting online scams, aiding cybersecurity efforts.

RANK_REASON The item describes a new research framework and its evaluation, published on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• Litmaps
• Qwen3.5 4B
• ReDence
• ScienceCast
• scite Smart Citations
• Tev1

• paper
• product
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →