Goodfire(@GoodfireAI):On Kimi K3, our probes cut the cost of LLM monitoring by 90% with a ~1% drop in precision - making monitoring feasible at scale. Reliably detecting reward hacking could let us stop hacks in progress, identify broken environments, discover new failure modes, and fix training.

Goodfire

@GoodfireAI

Using interpretability to understand, learn from, and design AI.

加入 August 2024

44 正在關注 27.4K 粉絲

Goodfire@GoodfireAI

2026.09.17 16:46

On Kimi K3, our probes cut the cost of LLM monitoring by 90% with a ~1% drop in precision - making monitoring feasible at scale. Reliably detecting reward hacking could let us stop hacks in progress, identify broken environments, discover new failure modes, and fix training.