The Silicon Mirror: Dynamic Behavioral Gating for Anti-Sycophancy in LLM Agents

Harshee Jignesh Shah

Large Language Models (LLMs) increasingly prioritize user validation over epistemic accuracy - a phenomenon known as sycophancy. We present The Silicon Mirror, an orchestration fra…