Jacques Thibodeau's repositories
map-floodwater-satellite-imagery
This repository focuses on training semantic segmentation models to predict the presence of floodwater for disaster prevention. Models were trained using SageMaker and Colab.
Weak-Supervised-Learning-Case-Study
Exploring NLP weak supervision approaches to train text classification models. The project is also a prototype for a semi-automated text data labelling platform. Approaches: Snorkel and Zero-Shot Learning.
ai-safety-scrape
Scraping different AI Safety resources.
aligning-language-models-mats
This is the repository for an 8-week research project that was worked on while attending SERI MATS.
gpt-experiments
This repository contains various experiments and prototypes to get use to working with GPT-like models and being creative with them.
anti-misinfo-helper
An AI-aided tool to add to Community Notes to improve efficiency and help people write notes.
supervising-ais-improving-ais
Future prosaic AIs will likely shape their own development or that of successor AIs. We're trying to make sure they don't go insane.
ai-safety-prize-challenge
A webapp for finding "bad" outputs of LLMs.
aligning-language-models
This repository contains experiments on aligning language models.
alignment-research-dataset
A dataset of alignment research and code to reproduce it
arxiv-alignment-paper-notifier
A tool to get all the latest AI alignment paers from arxiv.
cer-qa-app
This is an question-answering app for Canada Energy Regulator documents.
elk
Keeping language models honest by directly eliciting knowledge encoded in their activations. Building on "Discovering latent knowledge in language models without supervision" (Burns et al. 2022)
gpt-stackoverflow-QA
This repository focuses on fine-tuning GPT-J on StackOverflow questions and answers.
huggingface-course-cer-workshop
This is repository is for an abridged version of the Huggingface course on a Windows machine.
mesh-transformer-jax
Model parallel transformers in JAX and Haiku
mlab
Machine Learning for Alignment Bootcamp
orbyter-cookiecutter
Dockerized ML Cookiecutter
RLHI
Reinforcement Learning with Heuristic Imperatives - Finetuning LLMs for Post-Conventional Moral Intuition
rome-experiments
Locating and editing factual associations in pre-trained transformers
Software-Engineering-Best-Practices-for-Data-Scientists
A collection of the software engineering best practices for data scientists and ML engineers.
transformers-from-scratch
This is repository for learning about building transformer models from scratch.
white-box-rome
Using tuned lens to better understand the properties being projected at a specific layer-token.