4 terms
Showing all terms starting with J
A technique used to bypass an AI model's safety restrictions by crafting prompts that trick it into producing prohibited content.
A feature in some LLM APIs that constrains model output to valid JSON, making it easier to integrate AI responses into applications.
A technique where multiple modalities (text and images) are mapped into the same vector space, enabling cross-modal search and generation.
Compiler optimisation techniques applied to ML frameworks that speed up model inference by compiling computation graphs at runtime.