Nina

CAA. Steering Llama 2 with Contrastive Activation Addition

LM-exp. LLM experiments done during SERI MATS - focusing on activation steering / interpreting activation spaces

InfluenceFunctions. Implementation of Influence Function approximations for differently sized ML models, using PyTorch

devinterp. Quantifying degeneracy in toy models

ActivationDirectionAnalysis. Python

lmexp. Simple starter code for experiments on open-source LLMs. Built for my SPAR project participants.

mlexperiments. Exploratory work analyzing loss landscapes of neural nets / steering between generalizations / phase transitions

EmbedMap. Generate a 3D map of links based on their embeddings using OpenAI's embedding API

WritingStyleEmulation. Playing around with prompting techniques to get Claude to emulate other writing styles

BehaviorEvals. Scripts for evaluating LLMs

Feedbackr. Easily collect yes/no feedback on language model outputs from humans

city-cards. Python

RaccoonApp. Raccoon-themed habits app for iOS

InputColoring. role embeddings / input coloring project