Glossary entry
गार्डरेल / सुरक्षा सीमा
Definition (English)
Rules and safeguards that keep an AI system from producing harmful or unwanted output.
अर्थ (नेपालीमा)
एआई प्रणालीलाई हानिकारक वा अवाञ्छित नतिजा निकाल्नबाट रोक्ने नियम र सुरक्षा उपाय।
Related terms
Jailbreak
जेलब्रेक
Tricking an AI into bypassing its safety rules to produce restricted output.
Toxicity
विषाक्तता
Harmful, hateful, or abusive content that AI systems should detect and avoid producing.
AI Governance
एआई सुशासन
The rules, roles, and processes that guide responsible development and use of AI.
Autoregressive
अटोरिग्रेसिभ
Generating output one token at a time, each based on all the tokens produced before it.
Beam Search
बिम सर्च
A decoding method that keeps several candidate sequences to find a high-quality output.
Bias (parameter)
बायस (प्यारामिटर)
An extra learnable value added to a neuron's output to shift it as needed.
Conditioning
कन्डिसनिङ
Guiding a generative model's output using extra inputs like text, images, or labels.
Constraint Satisfaction
अवरोध सन्तुष्टि
Finding values for variables that satisfy a set of given rules or limits.