Large language models
S5 (1re) to S6 (Terminale) · 🤖 Artificial intelligence · 🛠️ Technology and digital skills
Tokens, attention, parameters and context window, pre-training and fine-tuning, limitations, safety and societal issues of LLMs.
- Questions
- 24, in 4 parts: How they work · Training · Limitations · Challenges
- Chapter
- Large language models
- Game
- Pixelia
- Pack language
- English
A taste of what is inside
- A large language model splits text into small pieces called...atomsoctavespixelstokens
- Pre-training a language model is done on...a single dictionaryimages onlyimmense text corporasounds
- Why can a model give a false yet convincing answer?it is brokenit is meanit lies on purposeit targets probability, not truth
How to get it
In your game (full version): ⚙️ Settings → 📚 Question bank, choose this pack, then “Buy”. The payment opens in your browser; the pack then installs itself and stays available on every device of your licence.
Try Pixelia: the free demo