Pinned Loading

1. evaluation-awareness-probing evaluation-awareness-probing Public

   probing and steering eval awareness

   Jupyter Notebook 7 8

2. probity probity Public

   Forked from curt-tigges/probity

   using whitebox monitors in oversight, e.g. debate

   HTML 1

3. location-inference location-inference Public

   models are superhuman at Geoguessr-style inference

   Jupyter Notebook

4. llm-self-preference llm-self-preference Public

   a system's preferences may be revealed by what it chooses to become

   Jupyter Notebook

5. replicating-belief-states-fractal replicating-belief-states-fractal Public

   https://www.lesswrong.com/posts/gTZ2SxesbHckJ3CkF/transformers-represent-belief-state-geometry-in-their

   Jupyter Notebook

6. introspective-models introspective-models Public

   Python