I would like to thank Jiří Materna & the Machine Learning Prague conference team.
They organized a blast conference in #Prague and I managed to speak for the 10 year edition.
A bit overload of #Llm but I guess it is the same everywhere

#mlconference #aiconference #machinelearning #AI #mlprague2025

I believe if anyone showed us GPT4.5 10 years ago, we would call it AGI.

(Standa Fort is killing it at #MLPrague2025)

Extending Context Window of Large Language Models via Positional Interpolation We present Position Interpolation (PI) that extends the context window sizes of RoPE-based pretrained LLMs such as LLaMA models to up to 32768 with minimal fine-tuning (within 1000 steps), while demon...

I had no idea that one of LLM innovations was invented for write better erotic stories. (arxiv.org/abs/2306.15595, enabling LLMs to work with longer texts)

What you learn at #mlprague2025!

Jon McLoone (Wolfram Research): People are asking AI chatbot wrong questions in the same way they were asking Google wrong questions 20 years ago. It is possible they will learn to do better.

#MLPrague2025

Low-background steel - Wikipedia

From questions to Jon McLoone's talk at ML Prague "Lies, Damn Lies and Gen AI"

Sunken WWII shipwrecks are valuable because they are source of low-background steel (en.wikipedia.org/wiki/Low-bac...). Maybe similarly, one day we will be seeking for pre-AI texts.

#MLPrague2025