rokosbasilisk

Scal­ing Laws for LLM Based Data Compression

Ex­plor­ing the Pla­tonic Rep­re­sen­ta­tion Hy­poth­e­sis Beyond In-Distri­bu­tion Data

Un­der­stand­ing Hid­den Com­pu­ta­tions in Chain-of-Thought Reasoning

Goal-mis­gen­er­al­iza­tion is ELK-hard

Hut­ter-Prize for Prompts

The AGI needs to be honest