- Attention Mechanism Attention lets a neural network weigh other elements by relevance when processing each one. It's the core computation inside Transformers and today's LLMs.
- Context Window The context window is the maximum number of tokens an LLM can process at once, which sets how much conversation, documents and instructions it can see.
- Diffusion Model Diffusion models generate data by learning to remove noise step by step. They power most of today's text-to-image and text-to-video tools. Here's how.
- Embedding An embedding turns text, images or other content into a vector of numbers, so computers can measure meaning by distance. It underpins search and RAG.
- Fine-tuning Fine-tuning continues training a pretrained model on task- or domain-specific data so it performs better at that work. Covers full fine-tuning, LoRA and more.
- Hallucination AI hallucination is when an LLM produces fluent, confident content that's false or made up. This entry covers what it is, why it happens and how to reduce it.
- Large Language Model (LLM) A large language model is a neural network trained on vast amounts of text to understand and generate language. This entry covers how LLMs are trained and used.
- Mixture of Experts (MoE) Mixture of Experts splits part of a network into many experts and activates only a few per input, growing parameter count while keeping compute in check.
- Multimodal AI Multimodal AI refers to models that can understand or generate several types of information, such as text, images, audio and video. Here's how they work.
- RLHF (Reinforcement Learning from Human Feedback) RLHF trains a reward model on human preference judgments, then uses reinforcement learning to make an LLM more helpful and safer. Covers steps and alternatives.
- Transformer The Transformer is a neural network architecture built around attention, introduced in the 2017 paper Attention Is All You Need. It underpins modern LLMs.
- Vector Database A vector database stores and searches high-dimensional vectors, quickly finding the most semantically similar items. It is core infrastructure for RAG.
Continua explorant què pot fer la IANavega per totes les eines