EXPLORE / 06
Terms index
Specialized AI and software-development terms with concise explanations
Entity records
4540 records · 108/114
- 4281
Explicit Breakpoints
Manually marking cache boundaries in prompts to precisely control cache scope.
Website pending - 4282
Frontier Model
State-of-the-art large foundation models with the most advanced general capabilities
Website pending - 4283
Multilingual Model
Pre-trained large models capable of understanding and generating multiple natural languages
Website pending - 4284
Human Oversight
Mechanisms for humans to monitor, intervene in, or control autonomous systems.
Website pending - 4285
AI Alignment
The field ensuring AI systems behave in accordance with human intent and values.
Website pending - 4286
Color Grading
The process of adjusting video tones to create a specific mood or visual style
Website pending - 4287
User Intent Recognition
Technology parsing natural language to clarify user goals
Website pending - 4288
Human-in-the-loop Control
A mechanism retaining human oversight and intervention capabilities within automated workflows.
Website pending - 4289
Proactive Agent
An AI system that autonomously initiates actions and plans multi-step tasks rather than merely responding to instructions.
Website pending - 4290
Misalignment
The phenomenon where an AI system's actual behavior or goals diverge from its intended design.
Website pending - 4291
Safety Case
A structured argument demonstrating that system risks are controlled and meet specific safety standards.
Website pending - 4292
API Token
The basic unit used to measure input and output text length in large model services.
Website pending - 4293
Adversarial Extraction
Stealing internal logic or data from black-box models via adversarial means
Website pending - 4294
Model Distillation
A technique training student models using teacher model outputs to compress knowledge
Website pending - 4295
Enterprise LLM
LLM versions offering security, compliance, and customization for business scenarios
Website pending - 4296
Reasoning Effort
A parameter controlling the depth of model thinking before generating an answer
Website pending - 4297
Frontier labs
Research organizations developing state-of-the-art large-scale AI models.
Website pending - 4298
Open-weight models
LLMs with publicly available parameters for download, modification, and deployment.
Website pending - 4299
Mathematical Reasoning
The ability of models to solve complex math problems via logical steps
Website pending - 4300
Trained from Scratch
Learning from data starting with random initialization without relying on pre-trained weights.
Website pending - 4301
Last-mile problem
Difficulties encountered in the final stage of a process or system deployment, preventing full success despite high initial performance.
Website pending - 4302
D&O Insurance
Insurance providing compensation for lawsuits against directors and officers arising from their official acts
Website pending - 4303
Multimodal
Models capable of processing and understanding multiple types of data inputs simultaneously.
Website pending - 4304
Vector Embedding
Converting unstructured data into numerical vectors for similarity computation.
Website pending - 4305
API Pricing
The fee structure developers pay based on usage when calling cloud services via Application Programming Interfaces
Website pending - 4306
Image Generation Model
A system using deep learning algorithms to create visual content from text prompts or other inputs
Website pending - 4307
Safety Alignment
Technical process ensuring AI system behaviors align with human values and safety guidelines
Website pending - 4308
Red Teaming
Adversarial testing method simulating attacker behavior to evaluate system security
Website pending - 4309
Self-Distillation
Process where a model extracts knowledge from its own or a teacher's intermediate representations to optimize itself.
Website pending - 4310
Rubrics
Detailed, structured textual guidelines used to evaluate the quality of model outputs.
Website pending - 4311
Guardrails
Constraints designed to limit AI outputs to prevent harmful or inappropriate behavior.
Website pending - 4312
Alignment
The process of ensuring AI systems behave in accordance with human values and intentions.
Website pending - 4313
Precision-Recall
Metrics pair measuring the accuracy and completeness of positive sample detection.
Website pending - 4314
Code Review
The process of examining source code before it is committed to find errors or improvements.
Website pending - 4315
Vector Similarity Search
Algorithm finding semantically similar content based on vector distance
Website pending - 4316
Agentic Retrieval
Mechanism where an AI agent plans and executes multi-step retrieval tasks
Website pending - 4317
Explainability
The degree to which a model's decision process is transparent and understandable to humans
Website pending - 4318
Idempotency
System property where multiple identical operations yield the same result
Website pending - 4319
Prompt Caching
Reusing computed prefix KV caches to reduce redundant calculations, lowering latency and cost.
Website pending - 4320
FedRAMP
A standardized process for assessing and authorizing cloud service providers' security by the US government.
Website pending