Search papers, labs, and topics across Lattice.
Plain-English definitions of the terms you'll encounter in AI research papers and discussions.
A large language model is an AI system trained on massive amounts of text to understand and generate human language. LLMs learn patterns from billions of words and can answer questions, write code, summarize documents, and reason through problems. Examples include GPT, Claude, Gemini, and Llama.
RLHF is a training technique where an AI model is fine-tuned using human preferences. Humans rank the model's outputs from best to worst, and this feedback trains a reward model that guides the AI toward more helpful, harmless, and honest responses. It's a key technique behind ChatGPT, Claude, and other conversational AI systems.
Computer vision is the field of AI that enables machines to interpret and understand visual information from images and videos. Applications include object detection, facial recognition, medical image analysis, autonomous driving, and image generation. Modern CV systems typically use deep neural networks.
A recurrent neural network is a type of neural network designed to process sequential data like text, audio, or time series. RNNs maintain a hidden state that acts as memory, allowing them to consider previous inputs when processing new ones. Variants like LSTMs and GRUs improved on basic RNNs, though transformers have largely replaced them for language tasks.
Multimodal AI refers to systems that can process and generate multiple types of data — such as text, images, audio, and video — within a single model. GPT-4o, Gemini, and Claude are multimodal models that can understand images alongside text. This contrasts with unimodal models that only handle one data type. See multimodal research on the Topics page.
The transformer is a neural network architecture introduced in the 2017 paper "Attention Is All You Need." It uses a mechanism called self-attention to process all parts of an input simultaneously rather than sequentially. Transformers are the foundation of virtually all modern LLMs, including GPT, Claude, Gemini, and Llama.
Natural language processing is the branch of AI focused on enabling computers to understand, interpret, and generate human language. NLP powers applications like translation, sentiment analysis, chatbots, search engines, and text summarization. Modern NLP is dominated by transformer-based language models.
A convolutional neural network is a type of deep learning model designed primarily for processing grid-like data such as images. CNNs use convolutional filters to automatically detect features like edges, textures, and shapes. They remain widely used in computer vision, though vision transformers (ViTs) are increasingly competitive.
Fine-tuning is the process of taking a pre-trained AI model and further training it on a smaller, specialized dataset to improve performance on a specific task. For example, a general-purpose LLM might be fine-tuned on medical literature to become better at answering health questions. It's more efficient than training a model from scratch.
A foundation model is a large AI model trained on broad data that can be adapted to many different tasks. GPT-4, Claude, and Llama are foundation models — they're trained once at great expense, then fine-tuned or prompted for specific uses. The term was coined by Stanford's HAI to distinguish these general-purpose models from task-specific ones.
RAG is a technique that combines a language model with an external knowledge retrieval system. Instead of relying solely on its training data, the model first searches a database or document collection for relevant information, then uses that context to generate a more accurate and up-to-date response. RAG reduces hallucinations and keeps AI responses grounded in real data.
An AI agent is a system that can autonomously plan, make decisions, and take actions to achieve a goal. Unlike a simple chatbot that responds to one prompt at a time, an agent can break down complex tasks, use tools (search, code execution, APIs), and iterate on its approach. Examples include coding agents, research agents, and browser-based agents. See the latest agent research on the Topics page.
A diffusion model is a type of generative AI that creates data (typically images) by learning to reverse a noise-adding process. During training, noise is gradually added to real images until they become pure static; the model learns to reverse each step. Stable Diffusion, DALL-E, and Midjourney all use diffusion models for image generation.
Prompt engineering is the practice of crafting effective instructions for AI models to get better outputs. This includes techniques like few-shot prompting (providing examples), chain-of-thought reasoning (asking the model to show its work), and system prompts (setting behavioral guidelines). Good prompts can dramatically improve AI performance without changing the model itself.
Transfer learning is a machine learning technique where a model trained on one task is reused as the starting point for a different task. For example, a model trained on millions of images can transfer its learned visual features to a medical imaging task with only a small amount of medical data. This makes it possible to build specialized AI systems without massive datasets for every use case.
Have more questions? Check out the FAQ.
Explore AI research topics in depth on the Topics page.
