Pinned
ML is so attractive to young physicists because many physics grads aren't really looking for physics, but for our times' great challenge. My generation still believed LHC was about to give us the secrets of the universe. We then found our own quantum mechanics in ML- attention sinks may be a bias in causal transformers. as some of you know, i've been writing a long blogpost on attention and its properties as a message-passing operation on graphs. while doing so, i figured i might have found an explanation for which attention sinks may be an
- performative woman activities>could’ve been a prostitute but i chose to run a llm locally on my foldable
- we can go beyond attention. as some of you know, higher-order attention methods (and the resulting schizodrawings) have been my focus for a while now, and, despite my earlier plans, they ended up being my choice for the second post in the series titled "the graph side of
- 2026 is gonna be the year of blogposts
- i am once again asking @wandb to make a phone app so that i can monitor the situation when i'm outside
- Replying to @scheminglunaticdoes it chat? Yes Is it a GPT? (most likely) yes also CLAUDE clearly stands for Command Line Ascetic Universial Development Engineer So your professor is based and your retarded
- arxiv.org/abs/2505.22785 is a quite underrated paper imo. generalizing means your latent trajectories are not collapsing to exactly a (compressed) carbon-copy of the input -> 0-dimensional (single point) manifold in latent space, but instead converging to an "attractor"
- every time i think of what a tokenizer is and how a transformer ingests indices and spits them out, i am reminded of how ghastly and unholy deep learning really is, despite how much i love it
- explanatory thread about the recent paper by @GladiaLabLLMs are injective and invertible. In our new paper, we show that different prompts always map to different embeddings, and this property can be used to recover input tokens from individual embeddings in latent space. (1/6)
- this is SF andrew tate, will not elaborate furtherCourse live now. $4999 for the next 24 hours cluely . university
