The process of ensuring that AI systems behave in ways that are consistent with human values and intentions.
Read More
The degree to which a model’s decision-making process can be understood by humans.
Designing prompts to test or exploit vulnerabilities in AI models.
An experimental, educational framework for exploring ergonomic, lightweight multi-agent orchestration.
A technique that allows models to focus on different parts of the input when generating output.
Chain-of-thought prompting is a strategy that encourages an AI model to articulate its reasoning process step-by-step. This method often leads to more accurate and transparent decision-making.
Techniques to align AI models with specific values or principles through careful prompt design.
Using prompts to limit the model’s output to specific formats or content types.
The maximum amount of text a model can process in a single prompt.
The ability of a model to apply knowledge from one type of prompt to a different but related task.
AI agents that can adapt their behavior and instructions based on context and previous interactions.
Dense vector representations of words, sentences, or other data types in a high-dimensional space.
AI systems designed to provide clear explanations for their outputs or decisions.
The process of selecting, modifying, or creating new features from raw data to improve the performance of machine learning models.
A machine learning technique that trains algorithms across multiple decentralized devices or servers holding local data samples.
Few-shot prompting is a method that involves providing a small number of examples to guide an AI model's performance on a task.
The process of further training a pre-trained model on a specific dataset to adapt it to a particular task or domain.
Enabling AI models to call specific functions to perform tasks or retrieve information.
A framework where two neural networks (a generator and a discriminator) compete against each other to create realistic data.
An optimization algorithm used to minimize the cost function in machine learning by iteratively updating the model parameters.
When an AI model generates false or nonsensical information that it presents as factual.
The model’s ability to adapt to new tasks based on information provided within the prompt.
Fine-tuning language models on datasets focused on instruction-following tasks.
A machine learning technique that aims to transfer the learnings of a large pre-trained model (the "teacher model") to a smaller "student model."
The date up to which an AI model has been trained on data, beyond which it doesn’t have direct knowledge.
LLMCompiler is a system that optimizes and parallelizes function calls for large language models, improving efficiency and performance compared to sequential approaches.
A compressed representation of data in which similar data points are closer together, often used in generative models.
A technique where complex tasks are broken down into simpler subtasks.
Using prompts that instruct the model on how to interpret or respond to subsequent prompts.
An open standard for connecting AI systems with data sources, enabling secure, two-way connections between data sources and AI-powered tools.
A technique for reducing model size by removing unimportant parameters from a neural network while maintaining performance.
A framework where multiple AI agents work together, each with specific roles and capabilities, to accomplish complex tasks.
Designing prompts that ask the model to perform multiple tasks simultaneously.
A field of AI that focuses on the interaction between computers and humans through natural language.
A set of algorithms inspired by the human brain that are designed to recognize patterns and process complex data inputs.
One-shot prompting is an approach where a single example is given to an AI model to inform its task execution. It's a minimal form of guidance that relies heavily on the model's ability to extrapolate from limited information.
When a model learns the training data too well, including its noise and peculiarities, leading to poor generalization on new data.
A method of improving the performance of pre-trained large language models by training a small set of parameters while preserving most of the model's structure.
AI systems designed to formulate a plan of action and then carry out that plan to achieve a specified goal.
A prompt is the input text given to an AI model to elicit a specific response or action. It serves as the primary means of communication between users and AI systems.
Enhancing prompts with additional context or information to improve performance.
Adjusting prompts to account for known biases or limitations of the model.
Connecting multiple prompts in a sequence to achieve more complex tasks.
Grouping similar prompts together to identify patterns or optimize prompt libraries.
The logical consistency and flow of information within a prompt.
Techniques to reduce prompt length while maintaining effectiveness.
Breaking down complex prompts into simpler, more manageable components.
The process of condensing longer, more complex prompts into shorter, more efficient versions while maintaining their effectiveness.
The practice of designing and optimizing prompts to achieve desired outcomes from AI models.
Using multiple different prompts and aggregating their results.
The specific structure and organization of information within a prompt.
Attempting to override the model’s intended behavior through carefully crafted prompts.
The process of refining and improving prompts based on the model’s outputs.
Unintended disclosure of sensitive information through carefully crafted prompts.
A collection of tested and effective prompts for various tasks.
Iteratively refining prompts to improve model performance on specific tasks.
Adding specific phrases or instructions at the beginning of a prompt to guide the model’s behavior.
The ability of a prompt to consistently produce desired outcomes across different inputs.
Gradually building up complexity in prompts to guide the model toward more sophisticated outputs.
The degree to which small changes in a prompt can affect the model’s output.
Systematically studying how small changes in prompts affect model outputs to understand robustness and behavior.
A reusable structure for creating effective prompts across different tasks.
Systematically evaluating the effectiveness of different prompts.
Removing unnecessary elements from a prompt to improve efficiency without sacrificing effectiveness.
Fine-tuning only a small set of task-specific prompt parameters while keeping the main model frozen.
Keeping track of different versions of prompts as they evolve.
A technique used to train language models based on human preferences and feedback.
A modular paradigm that detaches the reasoning process from external observations in Augmented Language Models, significantly reducing token consumption.
A type of machine learning where an agent learns to make decisions by taking actions in an environment to maximize cumulative reward.
Enhancing model responses by retrieving relevant information from external sources.
Assigning a specific role or persona to the AI model within the prompt to shape responses.
A method that generates multiple reasoning paths and selects the most consistent one.
Using AI to understand the meaning and context of search queries rather than just matching keywords.
AI agents with fixed instructions and predefined behaviors that don't change during execution.
A type of machine learning where the model is trained on labeled data, learning to map inputs to outputs.
A special type of prompt that sets the overall context or persona for the AI model.
Tailoring prompts for particular types of tasks such as summarization or translation.
A parameter that controls the randomness or creativity of the model’s output.
A variant of chain-of-thought prompting, focusing on maintaining coherent reasoning throughout a conversation or task.
The basic unit of text processed by a language model, often a word or part of a word.
A text generation method that samples from the most likely tokens, based on probability mass.
Applying knowledge gained from one task to improve performance on a different but related task.
A type of neural network architecture that uses self-attention mechanisms, commonly used in large language models.
When a model is too simple to capture the underlying patterns in the data, resulting in poor performance.
A type of machine learning that involves training a model on data without labeled outputs, focusing on finding patterns and structures.
Zero-shot prompting is a technique where an AI model is asked to perform a task without being provided any examples.
