5 ms·Detecting misbehavior in frontier reasoning models8 points by meetpateltech 2y agosiliconc0w 2y agoMakes you think China is gonna release an open weight LLM designed to ace benchmarks but secretly harbor "bad thoughts".