This is a heavily interactive web application, and JavaScript is required. Simple HTML interfaces are possible, but that is not what this is.
Post
~bc
bcooney.info
did:plc:ku7hcnzqzy36dcminsbqeeqw
super interesting read:
arxiv.org/abs/2512.08093
LLMs may bluff due to RL incentives. this paper adds an honesty-only “confession” after answers. trained models admit errors/hallucinations, enabling better monitoring and transparency.
https://arxiv.org/abs/2512.08093
2025-12-16T04:43:03.931Z