← home
Bookshelf
A running list of papers, books, and videos I’ve found interesting. Occasionally annotated, sometimes just worth reading.
Exploration of artificial intelligence, its future trajectory, and its implications for humanity.
Exploring how attention residuals influence model behavior and internal representations.
Post-training quantization method that preserves accuracy while enabling faster and more efficient inference, especially for large language models.
A simple modification to self-attention that excludes information from a token’s own value vector, encouraging stronger context modeling and improved long-sequence language modeling.
A visual explanation of entropy, information, and the connection between compression and intelligence.
A useful segment on self-play: models or agents improving by generating challenges, competing against themselves, and learning from the feedback loop.
NVIDIA's unified world-model framework for jointly processing and generating language, image, video, audio, and action sequences for physical AI systems.