Sheikh Abdur Raheem Ali's repositories
CAA
Steering Llama 2 with Contrastive Activation Addition
base-models-refuse
Code to reproduce key results accompanying "Base LLMs refuse too"
claude-code
Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.
crosscoder-model-diff-replication
Open source replication of Anthropic's Crosscoders for Model Diffing
devinterp
Quantifying degeneracy in toy models
dotfiles
my personal terminal configurations for alignment research engineering
fae
(jax, tpu) nf4 matmuls for flux + t5 + onnx vae. vision SAE training and maxacts
hoyolab-auto-daily
Easiest, full free, and no BS Hoyolab daily check-in using GitHub Actions. Supports Zenless Zone Zero, Honkai: Star Rail, Genshin Impact, Honkai Impact 3rd, and Tears of Themis.
Language-Model-SAEs
For OpenMOSS Mechanistic Interpretability Team's Sparse Autoencoder (SAE) research.
llm-viz
3D Visualization of an GPT-style LLM
marc
Public repository for "The Surprising Effectiveness of Test-Time Training for Abstract Reasoning"
motion-canvas
Visualize Your Ideas With Code
sae
Sparse autoencoders
sae-rm
Using SAE's to interpret Reward Models (RM)
SAE-TS
Improving Steering Vectors by Targeting Sparse Autoencoder Features
sapiens
High-resolution models for human tasks.
semantic-router
Superfast AI decision making and intelligent processing of multi-modal data.
sheikheddy.github.io
My personal website
slt
Tools for studying developmental interpretability in neural networks.
SwitchTransformers
Implementation of Switch Transformers from the paper: "Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity"
unit
Next Generation Visual Programming System