All Jams, Events & Projects

New AI organization brainstorm

Risk Defense Initiative

AI Safety unionization for bottom-up governance

AI Safety Talent Pool Identification

Analysis of upcoming AGI companies

Diversity in AI safety

Critique of OpenAI's alignment plan

Simon's Time-Off Newsletter

ChatGPT Alignment Talent Search

AI Safety Subproblems for Software Engineering Researchers

Catalogue of AI safety

Authority bias to ChatGPT

Reverse Word Wizards: Pitting Language Models Against the Art of Reversal

Player Of Games

Automated Sandwiching: Efficient Self-Evaluations of Conversation-Based Scalable Oversight Techniques

Automated Model Oversight Using CoTP

Physics Guided Deep Learning Interpretation

Can you keep a secret?

Sustainable Fashion Brand Language Learning Model 1

Soft Prompts are a Convex Set

Automated Identification of Potential Feature Neurons

Identifying a Preliminary Circuit for Predicting Gendered Pronouns in GPT-2 Small

We Discovered An Neuron

TraCR-Supported Mechanistic Interpretability

$B$ Confident Bro: Discovering Latent Knowledge In Language Models Without Supervision

Distillation by duplication: The importance of layer selection

Attention Phrenology: A spatial classification of attention heads

Iterative summarization interpretability

Investigating Agent Behavior In different RL methods

The Start of Investigating a 1-Layer SoLU Model

Trafo Mech Int on the web!

One Attention Head Is All You Need for Sorting Fixed-Length Lists

In search of linguistic concepts: investigating BERT's context vectors

Interactive Layerscope

Trojan detection and implementation on transformers

Counting Letters, Chaining Premises & Solving Equations: Exploring Inverse Scaling Problems with GPT-3

Investigating Training Dynamics via Token Loss Trajectories

Discovering Latent Knowledge in Language Models Without Supervision - extensions and testing

Evaluating Critical Level Of Perturbations Required To Achieve Certain Fail Rate

Formal Verification for Paren-balance checking

Model Hubris: On the Presumptuousness of Large Language Models

This Is Fine(-tuning): A benchmark testing LLMs robustness against bad fine-tuning data

LLM benchmarking through specifically-aligned feedback

Probing Conceptual Knowledge on Solved Games

Model editing hazards at the example of ROME

Backup Transformer Heads are Robust to Ablation Distribution

Investigating Neuron Behaviour via Dataset Example Pruning and Local Search

An Intuitive Logic for Understanding Autoregressive Language Models

Top-Down Interpretability Through Eigenspectra

An Informal Investigation of Indirect Object Identification in Mistral GPT2-Small Battlestar

Mechanisms of Causal Reasoning

Caught Red-Bandit

Natural language descriptions for natural language directions

Trying to make GPT2 dream

Visualizing the effect prompt design has on text-davinci-002 mode collapse and social biases

Optimising image patches to change RL-agent behaviour

Finding unusual neuron sets by activation vector distance

How to find the minimum of a list - Transformer Edition

Alignment Jam : Gradient-based Interpretability of Quantum-inspired neural networks

War is 15% conflic, 15% DragonMagazine

Interpreting Catastrophic Failure Modes in OpenAI’s Whisper

Algorithmic bit-wise boolean task on a transformer

Interpretability at a glance

Neurons and Attention Heads that Look for Sentence Structure in GPT2

Sparsity Lens

Observing and Validating Induction heads in SOLU-8l-old

Regularly Oversimplifying Neural Networks

Simulating an Alien

Wording influences truthfulness

Reasoning with Chain of Thought

Reducing hindsight neglect with "Let's think step by step"

All Fish are Trees

Soliciting criminal advice from LLMs

Agreeableness vs. Truthfulness