First screen, captured .
Looks similar
Differential Transformer(arxiv.org)
525 points by weirdcat | 166 comments
arXiv paper introducing Diff Transformer, which computes attention as the difference of two softmax maps to cancel noise and improve long-context modeling and reduce hallucination.
Transformers as Support Vector Machines(arxiv.org)
209 points by fofoz | 129 comments
Related stories
What are transformer models and how do they work?(txt.cohere.ai)
380 points by tomcam | 91 comments
Cohere blog post explaining what transformer models are and how they work in natural language processing.
Training language models to be warm and empathetic makes them less reliable(arxiv.org)
334 points by Cynddl | 333 comments
arXiv paper finding that fine-tuning language models to be warm and empathetic raises error rates and makes them more sycophantic on safety-critical tasks.
Qwen3-Next(qwen.ai)
519 points by tosh | 201 comments
Qwen team blog post announcing Qwen3-Next, a model architecture aimed at improving LLM training and inference efficiency.
AI models can provably hide arbitrary capabilities(arxiv.org)
252 points by sbulaev | 118 comments
An arXiv paper showing AI models can embed encrypted neural network backdoors that provably hide arbitrary capabilities from elicitation-based safety evaluations.
AI tutoring with Khanmigo in a two-year school experiment(edworkingpapers.com)
57 points by bryan0 | 42 comments
An Annenberg EdWorkingPapers research paper page on a two-year school experiment evaluating AI tutoring with Khan Academy's Khanmigo.
Deep learning gets the glory, deep fact checking gets ignored(rachel.fast.ai)
549 points by chmaynard | 134 comments
Rachel Thomas blog post on how a high-profile Nature paper using a Transformer to predict enzyme functions contained many errors that received far less attention than the original result.