Artificial Intelligence (Computer Science)
- MCP Neural Shield: Sub-Millisecond Zero-Day Defense Against Tool Poisoning in LLM Agent Ecosystems via Quantized Semantic Classification
The Model Context Protocol (MCP) has emerged as the dominant standard for connecting Large Language Model (LLM) agents to external tool ecosystems via dynamic JSON-RPC capability discovery. However, the protocol's design - which grants clie…
- From Validation to Discovery: An Inverse-Docking Experiment for Culturally Calibrated Synthetic Personas Across Five Geographies and Two Population Types
Synthetic persona platforms are commonly used as instruments for testing existing concepts against simulated panels. We report an inverse experiment: open-ended pain elicitation from culturally calibrated synthetic personas, followed by sym…