Unfaithful CoT transfers across model families, and resistance isn't reasoning
Inkhaven day 5: why doesn't the big model just eat the small model?
God I Love AIslop
Day 4 staying up until 4am to write some slop
Personas as pointers
Using other people's vibes as shortcuts to states of mind
Reasoning Model Activations May Encode Hint Influence That Chain-of-Thought Doesn't Reveal
Linear probes detect hint influence in OLMo reasoning model activations
What are thinking traces even for?
Inkhaven day 1
Computational description of a Diels-Alder reaction
Doing some DFT calculations