A reference from Weave

LLM fundamentals glossary

Build a practical understanding of language models, prompts, tokens, and context. Focus on the concepts that affect real applications and engineering workflows.

All terms

148 terms

  • Prompt engineering

    Prompt engineering is the practice of designing and testing instructions, examples, context, and output requirements that guide a language model toward a useful result. It is an iterative engineering activity supported by evaluation and observability.

    LLM fundamentals

  • Prompt injection

    Prompt injection is an attack or unintended instruction that causes a language model to treat untrusted content as higher-priority guidance. It can redirect an application, expose data, or trigger tools in ways the developer did not intend.

    LLM fundamentals

  • Prompt injection defense

    Prompt injection defense is a language-model concept about instruction design and control. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Prompt registry

    Prompt registry is a language-model concept about representation and similarity. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Prompt robustness

    Prompt robustness is a language-model concept about representation and similarity. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Prompt template

    Prompt template is a language-model concept about generation behavior and sampling. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Prompt variable

    Prompt variable is a language-model concept about context selection and limits. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Prompt version

    Prompt version is a language-model concept about instruction design and control. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Proximal policy optimization

    Proximal policy optimization is a language-model concept about serving behavior and operational tradeoffs. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Random seed

    Random seed is a language-model concept about context selection and limits. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Reasoning trace

    Reasoning trace is a language-model concept about representation and similarity. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Reinforcement learning from AI feedback

    Reinforcement learning from AI feedback is a language-model concept about training behavior and measurement. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Relative positional encoding

    Relative positional encoding is a language-model concept about serving behavior and operational tradeoffs. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Repetition control

    Repetition control is a language-model concept about generation behavior and sampling. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Residual connection

    Residual connection is a language-model concept about serving behavior and operational tradeoffs. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Response prefilling

    Response prefilling is a language-model concept about representation and similarity. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Retrieval augmented generation

    Retrieval augmented generation, or RAG, is an architecture that retrieves relevant documents or records and supplies them to a language model as context for generating a response. It can ground answers in a changing or private information source without retraining the model for every update.

    LLM fundamentals

  • Reward model

    Reward model is a language-model concept about mechanism and information flow. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Rotary positional embedding

    Rotary positional embedding is a language-model concept about mechanism and information flow. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Self-consistency decoding

    Self-consistency decoding is a language-model concept about context selection and limits. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Self-critique prompting

    Self-critique prompting is a language-model concept about generation behavior and sampling. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Sentence representation

    Sentence representation is a language-model concept about context selection and limits. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Sparse autoencoder

    Sparse autoencoder is a language-model concept about evaluation design and failure analysis. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Special token

    Special token is a language-model concept about representation and similarity. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Step-back prompting

    Step-back prompting is a language-model concept about generation behavior and sampling. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Structured output

    Structured output is a model response constrained to a defined shape such as JSON with named fields and types. It gives application code a predictable contract while leaving the model responsible for generating the field values.

    LLM fundamentals

  • Superposition

    Superposition is a language-model concept about mechanism and information flow. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Task decomposition prompting

    Task decomposition prompting is a language-model concept about representation and similarity. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Temperature

    Temperature is a generation parameter that changes how a language model samples from the probability distribution of possible next tokens. Higher values generally allow more variation, while lower values generally favor higher-probability choices.

    LLM fundamentals

  • Temperature scaling

    Temperature scaling is a language-model concept about instruction design and control. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Token embedding

    Token embedding is a language-model concept about generation behavior and sampling. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Tokenization

    Tokenization is the process of converting text into the token units a language model receives and generates. A token can represent a word, part of a word, punctuation, or another piece of text, depending on the tokenizer.

    LLM fundamentals

  • Top-k sampling

    Top-k sampling is a language-model concept about representation and similarity. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Top-p sampling

    Top-p sampling, also called nucleus sampling, selects the next token from the smallest group whose cumulative probability reaches a chosen probability p. Lower p values restrict the candidate set, while higher values allow more candidates.

    LLM fundamentals

  • Training batch size

    Batch size is a language-model concept about serving behavior and operational tradeoffs. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Training data contamination

    Training data contamination is a language-model concept about training behavior and measurement. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Training loss

    Training loss is a language-model concept about training behavior and measurement. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Training step

    Training step is a language-model concept about training behavior and measurement. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Transformer block

    Transformer block is a language-model concept about mechanism and information flow. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Tree of thoughts

    Tree of thoughts is a language-model concept about instruction design and control. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Typical sampling

    Typical sampling is a language-model concept about context selection and limits. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Underfitting

    Underfitting is a language-model concept about mechanism and information flow. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Validation loss

    Validation loss is a language-model concept about serving behavior and operational tradeoffs. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Vector index

    Vector index is a language-model concept about instruction design and control. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Vector normalization

    Vector normalization is a language-model concept about context selection and limits. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Vector similarity

    Vector similarity is a language-model concept about representation and similarity. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Warmup steps

    Warmup steps is a language-model concept about mechanism and information flow. It names a mechanism, representation, training practice, or operational behavior that can change how an AI system processes input and produces output.

    LLM fundamentals

  • Zero-shot prompting

    Zero-shot prompting is a language-model concept about instruction design and control. It names a mechanism, representation, prompting pattern, decoding control, or context behavior that can change how an AI system processes input and produces output.

    LLM fundamentals