[{"id":"embedding","term":"Embedding","aliases":["embeddings","vector embedding","embedding vector"],"short":"A list of numbers that represents the meaning of a piece of data, so that similar things end up with similar numbers.","category":"llms","categoryName":"LLMs"},{"id":"fine-tuning","term":"Fine-tuning","aliases":["fine tuning","finetuning"],"short":"Training an already-trained model a little more on a smaller, focused dataset so it gets better at a specific task or style.","category":"training","categoryName":"Training"},{"id":"hallucination","term":"Hallucination","aliases":["confabulation","AI hallucination"],"short":"When an AI model produces information that sounds confident and plausible but is false or made up.","category":"llms","categoryName":"LLMs"},{"id":"large-language-model","term":"Large language model","aliases":["LLM","language model"],"short":"A very large neural network trained on huge amounts of text to predict the next piece of text, which lets it write, summarize, and answer questions.","category":"llms","categoryName":"LLMs"},{"id":"machine-learning","term":"Machine learning","aliases":["ML"],"short":"A way of building software that learns patterns from examples instead of following rules written by hand.","category":"basics","categoryName":"Basics"},{"id":"neural-network","term":"Neural network","aliases":["artificial neural network","ANN","neural net"],"short":"A model made of layers of simple connected units that pass numbers forward, loosely inspired by neurons in the brain.","category":"basics","categoryName":"Basics"},{"id":"overfitting","term":"Overfitting","aliases":["overfit","over-fitting"],"short":"When a model memorizes its training data instead of learning the general pattern, so it scores well in training but poorly on new data.","category":"training","categoryName":"Training"},{"id":"oversampling","term":"Oversampling","aliases":["over-sampling","upsampling"],"short":"Repeating or creating more examples of a rare class so a dataset is more balanced before training.","category":"data","categoryName":"Data"},{"id":"prompt-injection","term":"Prompt injection","aliases":["indirect prompt injection","jailbreak"],"short":"An attack where someone hides instructions in text a language model reads, so the model follows the attacker's instructions instead of the user's.","category":"safety","categoryName":"Safety"},{"id":"regularization","term":"Regularization","aliases":["regularisation","L1 regularization","L2 regularization","weight decay"],"short":"Techniques that discourage a model from becoming too complex during training, which helps it avoid overfitting.","category":"training","categoryName":"Training"},{"id":"rag","term":"Retrieval-augmented generation","aliases":["RAG","retrieval augmented generation"],"short":"A technique where a language model first looks up relevant documents, then writes its answer using them, so answers can be grounded in real sources.","category":"llms","categoryName":"LLMs"},{"id":"token","term":"Token","aliases":["tokens","tokenization","tokenizer"],"short":"The small chunk of text a language model reads and writes, often a word, part of a word, or a punctuation mark.","category":"llms","categoryName":"LLMs"},{"id":"transformer","term":"Transformer","aliases":["transformer model","transformer architecture"],"short":"A neural network design that uses attention to weigh how every part of the input relates to every other part. Most modern language models are built on it.","category":"llms","categoryName":"LLMs"},{"id":"underfitting","term":"Underfitting","aliases":["underfit"],"short":"When a model is too simple or trained too little to capture the pattern in the data, so it does poorly even on its training data.","category":"training","categoryName":"Training"},{"id":"validation-set","term":"Validation set","aliases":["validation data","dev set","holdout set"],"short":"A slice of data kept out of training and used to check the model and tune its settings while you build it.","category":"data","categoryName":"Data"}]