I just shared a new article, "The State of Reasoning Models", where I am exploring 12 new research articles on improving the reasoning capabilities of LLMs (all published after the release of DeepSeek R1): https://
magazine.sebastianraschka.com/p/state-of-llm
-reasoning-and-inference-scaling
… 1. S1: Simple test-time scaling
2. Test-Time
@rasbt
-
State of Reasoning Models: 12 Research Articles on LLM Improvements
By
–
-
PyTorch GPU CUDA Auto-Detection on Machine Installation
By
–
On a non-CUDA machine it will actually get the CPU version automatically. But yeah, if you are on a CUDA machine, I guess it's assumed you want the CUDA version (makes sense though, because otherwise new users might be confused why their GPU is not working in PyTorch)
-
PyTorch installation size reaches 2.8 GB with uv pip
By
–
Fun fact, `(uv) pip install torch` is 2.8 Gb now 2
-
Apple Intelligence Models Limitations Despite iPhone Integration
By
–
Apple uses it on their phones, but yeah, Apple Intelligence models are also a bit limited
-
Exciting inference-scaling techniques and future AI developments
By
–
there are lots of cool inference-scaling techniques out there! More to come in my next Ahead of AI post
-
Key Insights from Annual Machine Learning Competitions Report
By
–
Every Spring, I'm excited to read through the comprehensive ~100-page "State of Machine Learning Competitions" report, which offers many interesting insights into current trends, useful tools, and emerging methodologies in the field. Below are some key takeaways from the latest
-
OpenAI’s Reasoning Approach: Response Generation as Better Term
By
–
It's an OpenAI thing (
https://
openai.com/index/learning
-to-reason-with-llms/
…). I think "Response generation" would probably be a better, more general term. -
Focus on Writing About Reasoning Models Over Video Content
By
–
Good question. Currently, I am more focused on writing about reasoning models (vs making videos on reasoning models) because the research landscape if evolving so quickly. Probably I'll make one in the future (more likely 2026ish) once the methodology converged.
-
End-to-End Encryption vs AI Computing Server Requirements
By
–
Interesting. I am surprised that they need it for AI computing but not for iCloud storage. But maybe because the data is send and stays end-to-end-encrypted whereas with AI computations you have to at least temporarily decrypt it on the server to run the query & computation.
-
AI-Generated Content Cannot Replace Authentic Human Creators
By
–
100%. I also don't even like AI-generated audiobooks or podcasts. I just can't listen to them. That's why, for me, they will never replace awesome creators & channels like @3blue1brown
. (The only thing is maybe they make the work easier / enable even higher quality videos for