EXPLORE / 06
Terms index
Specialized AI and software-development terms with concise explanations
Entity records
4517 records · 76/113
- 3001
TGI
A production-grade LLM inference server by Hugging Face designed for efficient deployment.
Website pending - 3002
VQA
Combines computer vision and NLP to answer questions about image content
Website pending - 3003
Multilinguality
The capability to understand and generate text across multiple natural languages
Website pending - 3004
Diffusion Transformer (DiT)
A technique applying Transformer architecture as the backbone of diffusion models
Website pending - 3005
Model Inference
The process of using a trained model to make predictions on new data
Website pending - 3006
Serverless
A cloud computing execution model where the provider dynamically manages allocation of machine resources
Website pending - 3007
Data Augmentation
Technique that generates new samples by transforming original data to improve model generalization
Website pending - 3008
Data Version Control
Technology for tracking and managing datasets similar to how Git manages code.
Website pending - 3009
Model Deployment
The process of integrating trained models into production environments for actual use
Website pending - 3010
Infini-Attention
An attention mechanism achieving linear complexity for long sequences via compressed historical states
Website pending - 3011
Data Packing
Concatenating multiple short sequences into fixed-length long sequences to reduce padding waste
Website pending - 3012
Hugging Face
A community and company hosting open-source models and datasets
Website pending - 3013
Video Encoding
Process of converting raw video into compressed formats
Website pending - 3014
Mixed Precision
Combining different floating-point precisions to optimize memory and speed
Website pending - 3015
Data Deduplication
Technology that identifies and eliminates duplicate copies of data to save storage space
Website pending - 3016
Parquet
A columnar binary file format commonly used for big data storage and analytics
Website pending - 3017
Server CPU
High-core-count, reliable central processing units designed for data centers
Website pending - 3018
Gradient Accumulation
A technique that accumulates gradients over multiple forward/backward passes to simulate larger batch sizes
Website pending - 3019
Keras
A high-level neural networks API written in Python, often used for rapid prototyping
Website pending - 3020
Parser
Program component for validating and transforming text structures, here referring to constrained decoding logic
Website pending - 3021
Structured Generation
Technique constraining LLM outputs to strictly follow specific formats like JSON or regex
Website pending - 3022
WASM
A binary instruction format enabling efficient execution of non-JavaScript code in browsers.
Website pending - 3023
Static Analysis
Method of checking source code or binaries for vulnerabilities without executing them
Website pending - 3024
Model Security
Measures protecting ML models from tampering, reverse engineering, or adversarial attacks
Website pending - 3025
Long Video QA
Question answering tasks targeting videos with longer durations and complex temporal information
Website pending - 3026
Adversarial Dataset Refinement
A method using adversarial mechanisms or strong models to generate distractors for filtering and optimizing dataset quality
Website pending - 3027
Zero-configuration
Mechanism that automates deployment and optimization without manual parameter setup.
Website pending - 3028
Sampling Strategy
Methods determining how LLMs select the next token from probability distributions.
Website pending - 3029
Multilingual LLM
Large language models capable of processing multiple natural languages
Website pending - 3030
Assistant Model
A smaller model that provides preliminary predictions or drafts during the main model's inference process.
Website pending - 3031
Open Dataset
Machine learning data collections freely accessible, usable, and modifiable by the public
Website pending - 3032
Human Preference Alignment
Training process that aligns model behavior with human values and preferences.
Website pending - 3033
Early Exit
A strategy allowing data to yield results at shallower network layers, thereby reducing computation and accelerating inference.
Website pending - 3034
Self-Attention
Mechanism allowing parallel processing and focusing on relationships between parts
Website pending - 3035
Positional Encoding
Vector representation method injecting sequence order info into Transformers
Website pending - 3036
Multi-Agent Debate
A technique where multiple AI agents improve answer quality by questioning and arguing with each other.
Website pending - 3037
Data Labeling
The process of adding tags or annotations to raw data for supervised learning purposes.
Website pending - 3038
Named Entity Recognition (NER)
A technique that identifies and classifies named entities like persons, locations, and organizations in text.
Website pending - 3039
Pretrained Weights
The set of parameters obtained after training a model on large-scale datasets, used for fine-tuning downstream tasks
Website pending - 3040
SigLIP
An improved visual encoding architecture based on sigmoid loss function, derived from CLIP
Website pending