EXPLORE / 06
Terms index
Specialized AI and software-development terms with concise explanations
Entity records
4377 records · 26/110
- 1001
Over-The-Air Update
Remote updating of device software or model parameters via network, enabling upgrades and fixes without physical connection.
Website pending - 1002
On-Device Model Compression
Techniques like quantization and pruning that reduce model size for efficient execution on resource-constrained mobile or embedd
Website pending - 1003
Motion Control
The planning and management of actuator sequences (joints, motors) to ensure stable and precise trajectory or posture adjustment
Website pending - 1004
Distillation Attack
In security contexts, often refers to extracting sensitive data or model behaviors indirectly; here specifically denotes targete
Website pending - 1005
Edge Inference Memory (EIM)
A storage architecture stacking low-power memory on logic chips to reduce data movement latency and increase bandwidth for edge
Website pending - 1006
Superchip
A high-performance module integrating CPU, GPU, and fast interconnects on a single substrate via advanced packaging for heteroge
Website pending - 1007
Context Pollution
The phenomenon where irrelevant or erroneous historical information mixes into the current context window, causing attention dis
Website pending - 1008
Agent Operating System
A software layer providing a unified runtime for AI agents, managing memory, tool invocation, task scheduling, and model routing
Website pending - 1009
Apache Parquet
A columnar storage format optimized for analytics, enabling efficient compression and encoding for fast data processing in big d
Website pending - 1010
BM25
A probabilistic ranking function used in information retrieval to estimate the relevance of documents to a given search query.
Website pending - 1011
Anti-Automation Detection
Techniques used by websites to identify and block non-human traffic (including scrapers and AI agents), often based on behaviora
Website pending - 1012
PCA Whitening
A data preprocessing technique using Principal Component Analysis to decorrelate features and normalize variance, eliminating in
Website pending - 1013
Agent-Native
A design paradigm where software or content structure and interaction are built around autonomous agent capabilities rather than
Website pending - 1014
Industrial Software
Specialized software for design, simulation, manufacturing, and management (e.g., CAD/CAE/CAM), characterized by high closedness
Website pending - 1015
Reinforcement Learning with Self-Feedback (RLSF)
A post-training method that uses the model's own internal confidence scores or self-generated critiques as reward signals, elimi
Website pending - 1016
Pareto Improvement
A change that improves at least one objective without worsening any other, often used to describe simultaneous gains in performa
Website pending - 1017
Code Porting
The process of migrating software from one platform, architecture, or programming language to another, often involving dependenc
Website pending - 1018
AI4S
A paradigm using deep learning to accelerate scientific discovery by processing massive datasets and simulating complex physical
Website pending - 1019
Activation Checkpointing
A technique that saves memory by discarding intermediate activations during forward pass and recomputing them during backward pa
Website pending - 1020
Open Banking
An architecture using APIs to allow third-party providers secure access to bank data with user consent, fostering financial serv
Website pending - 1021
Agent Interconnect
The architecture or mechanism enabling multiple autonomous agents to communicate, collaborate, synchronize states, and distribut
Website pending - 1022
Deterministic Networking
A networking technology that guarantees data transmission within specific time windows with extremely low and predictable latenc
Website pending - 1023
Knowledge Work
Tasks involving information processing, analysis, and professional judgment (e.g., finance, law), requiring long-context underst
Website pending - 1024
AI Skills
Predefined atomic task units encapsulating domain-specific prompts, tool-calling logic, and workflows, enabling general LLMs to
Website pending - 1025
CuTe DSL
A C++ template-based domain-specific language in CUTLASS for abstracting and writing efficient low-level GPU tensor core operati
Website pending - 1026
Visual Agent
An autonomous AI system that perceives visual inputs and executes complex operations (e.g., software interaction, 3D modeling),
Website pending - 1027
LOESS Smoothing
A local weighted regression method used to reduce noise while preserving trends in time-series visualization.
Website pending - 1028
System on Chip
An integrated circuit combining CPU, GPU, NPU, and memory controllers into a single package, suitable for high-efficiency embedd
Website pending - 1029
PRNG
A deterministic algorithm producing pseudo-random sequences dependent on an initial seed, commonly used for generating cryptogra
Website pending - 1030
Instruction Cache Locality
The arrangement of frequently executed code closely in memory to reduce latency when loading instructions from main memory, impr
Website pending - 1031
PGO
A compiler optimization technique that collects execution data from representative workloads to guide code layout, inlining, and
Website pending - 1032
Continuous Quality Loop
An engineering paradigm embedding quality assurance into the full lifecycle via real-time monitoring, automated feedback, and it
Website pending - 1033
Task Routing
A mechanism that dynamically assigns requests to different models or processing nodes based on task complexity, reasoning requir
Website pending - 1034
Agent Context Management
The process of integrating, retrieving, and structuring information from multi-source systems to provide background knowledge re
Website pending - 1035
MLP
Feedforward neural network composed of multiple fully connected layers; in Transformers, it serves as the position-wise feed-for
Website pending - 1036
In-silico Design
Using computer simulations and algorithms to design and screen molecular structures virtually, reducing physical experimentation
Website pending - 1037
Vulkan Backend
A compute acceleration layer based on the cross-platform Vulkan API, enabling llama.cpp to perform tensor operations on non-NVID
Website pending - 1038
GGUF Format
A binary file format designed for llama.cpp to store quantized LLM weights and metadata, optimizing hybrid CPU/GPU inference eff
Website pending - 1039
Automated Evaluator
A programmatic system that automatically checks and scores generated outputs against predefined metrics (e.g., gate count, fidel
Website pending - 1040
Quantum Circuit
A sequence of quantum gates acting on qubits to execute algorithms; depth and width determine hardware resource consumption.
Website pending