Słownik AI
Kompletny słownik sztucznej inteligencji
Common Sense Reasoning
Ability of a system to infer logical conclusions from non-explicit but universally understood information by humans to solve complex problems.
Implicit Inference
Process of deducing information that is not directly stated in the input text but is necessary to provide a complete and relevant answer.
Common Sense Knowledge Base
Organized data structure containing facts, rules, and relationships about the real world, used to power the reasoning of a question-answering system.
Reasoning Grid
Methodology or framework that breaks down a complex question into simpler reasoning sub-problems, often using common sense knowledge to link the steps.
Chain-of-Thought
Prompting technique that encourages an AI model to generate the intermediate steps of its reasoning, thereby improving its ability to solve logical and arithmetic problems.
Semantic Coherence Verification
Mechanism that ensures the generated answer is logically compatible with common sense knowledge and the context provided in the initial question.
Implicit Knowledge Generation
Ability of a model to create and articulate common sense premises that, although absent from the original text, are essential to justify an answer.
Common Sense Augmented Language Model
Language model architecture integrated or coupled with a common sense knowledge base, specifically designed to improve performance in implicit reasoning.
Reasoning Elicitation
Set of techniques aimed at extracting and making explicit the internal reasoning process of an AI model, often to make it more interpretable and reliable.
Commonsense Knowledge Graphs
Structured representation of commonsense knowledge in the form of nodes (entities) and edges (relations), facilitating logical paths for reasoning.
Logical Abduction
Form of reasoning that consists of inferring the most plausible hypothesis that can explain a set of observations, widely used to interpret ambiguous situations.
Qualitative Reasoning
Branch of reasoning that focuses on the symbolic and non-numerical aspects of a problem, manipulating concepts like 'larger', 'hotter', or 'before'.
Premise Validation
Step in the QA process that checks if the premises (explicit or implicit) on which a question rests are true, false, or undetermined according to commonsense knowledge.
External Knowledge Integration
Process by which a QA system dynamically merges information from external sources (like commonsense bases) with the question context to enrich its reasoning.
Reasoning about Causes and Consequences
Application of causal logic to answer questions that link an event to its origins or its future effects, requiring a deep understanding of real-world relationships.