Pinned Loading

1. unlearning_evaluation unlearning_evaluation Public

   Code for the paper "Do Unlearning Methods Remove Information from Language Model Weights?"

   Python 11 2

2. reasoning_or_performing reasoning_or_performing Public

   checking if reasoning LLMs use the reasoning they provide or they just output it because they're trained to

   Jupyter Notebook 1

3. david-dye/Ankh david-dye/Ankh Public

   Domain-specific language (DSL) for writing cryptographically-secure programs robust to timing attacks and memory leaks.

   C++

4. llm_probe llm_probe Public

   linear probes are used to make classifications based on activations, what happens if you use another llm as a classifier?

   Python 1

5. transformer_from_scratch transformer_from_scratch Public

   Building a transformer using only pytorch documentation and Vaswani et al., 2017 as reference (though edits from memory made to make it a GPT)

   Jupyter Notebook 1

6. reward_seeker reward_seeker Public

   Python 1 1