Tag: LLM Architecture

1. An NES CPU compiled into Qwen weights, no training required (github.com)
   AI · 3 days ago · October 31, 2026
2. No, looped transformers are not hiding the model's reasoning (magazine.sebastianraschka.com)
   AI · 1 month ago · September 10, 2026
3. How models learn to think harder, or less (magazine.sebastianraschka.com)
   Engineering · 3 months ago · July 26, 2026
4. Where a hybrid model beats a transformer, token by token (allenai.org)
   AI · 4 months ago · June 28, 2026
5. A new paper gives language models a sleep phase (arxiv.org)
   AI · 5 months ago · May 26, 2026
6. How 2026 open models buy long-context efficiency without shrinking (magazine.sebastianraschka.com)
   Engineering · 5 months ago · May 16, 2026
7. Ai2's EMO trains a mixture of experts you can run at one-eighth size (huggingface.co)
   AI · 5 months ago · May 8, 2026
8. DeepSeek-V4 spends most of its design budget making long context usable (huggingface.co)
   AI · 6 months ago · April 24, 2026
9. A field guide to the attention variants modern LLMs actually use (magazine.sebastianraschka.com)
   Engineering · 7 months ago · March 22, 2026
10. Sebastian Raschka's map of inference-time scaling for reasoning (magazine.sebastianraschka.com)
    Engineering · 9 months ago · January 24, 2026

← all tags