Get Started
Home
Topics
Search
Library
Research questionHow can smaller models reliably detect risk in obfuscated Chinese SMS under low-latency constraints?Obfuscation can hide risk-bearing phrases while keeping messages readable to people, causing direct classifiers to miss fraudulent content. Production systems must handle these evasive messages without exceeding practical latency or deployment-size limits.
AI
Machine Learning
Natural Language Processing
Research Paper
Small / On-device Models
Technology
Latest papersRecent research connected to this question, newest first.ReCAST: Restoration-aware Cascaded Stage-wise Training for Obfuscated SMS Risk ClassificationThe source studies restoration-aware cascaded training and teacher-to-student distillation for Chinese SMS risk classification. Evidence comes from an internally constructed real-world Chinese SMS benchmark and supports claims about obfuscation robustness and smaller deployable models within that setting.research paper · Sep 4, 2026
Related questions
How can diffusion language models support reliable mobile-edge agents under tight latency and resource constraints?How can multilingual NLP models reliably detect harmful posts and verifiable factual claims in low-resource languages?How can network operators detect risky LLM-generated configurations before deployment, even when translation accuracy is high?Can text-to-image models match web-scale performance using smaller, reproducible datasets and models?