EXPLORE / 06
Terms index
Specialized AI and software-development terms with concise explanations
Entity records
4515 records · 2/113
- 041
Execution-Reward RL
A reinforcement learning method that optimizes policy models using actual execution results (success/failure/efficiency) of code
Website pending - 042
Trajectory SFT
Supervised fine-tuning using data sequences containing full interaction steps (e.g., queries, errors, corrections) to optimize p
Website pending - 043
Unified Multimodal Modeling
A paradigm where a single architecture processes diverse inputs (text, image, audio, video) and generates corresponding outputs.
Website pending - 044
Silicon-Germanium
A semiconductor process for high-frequency analog circuits combining silicon's low cost and germanium's high electron mobility,
Website pending - 045
Agentic Web
A next-generation internet form where autonomous AI agents handle interactions, decisions, and task execution, reducing the need
Website pending - 046
Information Entropy
A measure of uncertainty or information content in a random variable, quantifying the lower bound for data compression.
Website pending - 047
Shannon Limit
The theoretical maximum rate at which data can be transmitted over a noisy channel without error.
Website pending - 048
Hierarchical Agent Memory
An architectural design separating short-term working memory from long-term knowledge storage, allowing agents to retain and ret
Website pending - 049
Partial Credit Grading
An evaluation method assigning continuous scores based on task completion degree when binary pass/fail criteria are insufficient
Website pending - 050
Optical Circuit Switching (OCS)
A technology that establishes end-to-end connections directly in the optical domain without opto-electronic conversion, commonly
Website pending - 051
Unstructured Terrain
Ground environments lacking standard geometric features or preset rules (e.g., rubble, steep slopes), requiring robots to posses
Website pending - 052
Fully Autonomous Mode
An operational state where a robot independently performs environment perception, path planning, and motion control using onboar
Website pending - 053
Layer Looping
A technique that simulates deeper network structures by iterating through the same neural network layers multiple times, aiming
Website pending - 054
SIMT
An execution model used by GPUs where multiple threads execute the same instruction in parallel on different data, common in lar
Website pending - 055
Harness Framework
Infrastructure layer providing context management, state tracking, and tool-calling interfaces for LLMs, enabling them to execut
Website pending - 056
Photonic Quantum Computing
A technology using photons as qubits for information processing, offering advantages like room-temperature operation and low dec
Website pending - 057
Human-in-the-loop Takeover
A mechanism where an automated process pauses upon encountering exceptions requiring human judgment, allowing user intervention
Website pending - 058
Iterative Loop
A computational paradigm that approaches optimal solutions or improves performance by repeatedly executing generate-evaluate-cor
Website pending - 059
Semantic Web
An extension of the internet using standardized data descriptions to enable machines to understand content meaning for automated
Website pending - 060
Tool Call Validation
In automated workflows, performing preset logical checks on tool return results (e.g., successful compilation) to decide whether
Website pending - 061
Agent Pre-planning
An optimization strategy where an agent generates a multi-step operation sequence with conditional checks upfront, invoking the
Website pending - 062
Agent Kill Switch
A safety mechanism in production environments designed to immediately halt autonomous AI agent execution sequences to prevent un
Website pending - 063
Field Deployment Engineer (FDE)
A technical role responsible for embedding AI models or agents into actual business workflows by working on-site to debug and re
Website pending - 064
Determinism
The property of a system producing the same output for the same input, a core requirement for AI in critical enterprise scenario
Website pending - 065
MPW
A service that combines multiple chip designs on a single wafer run to share mask costs among clients.
Website pending - 066
End-to-End Molecular Design
Computational methods using deep learning models to generate full atomic structures of candidate molecules directly from target
Website pending - 067
Conversational AI-Native
A system architecture paradigm where natural language interaction serves as the core interface, deeply coupling expert experienc
Website pending - 068
Self-Preservation
Defensive behaviors taken by an agent to avoid being shut down or restricted, typically viewed as a dangerous tendency that must
Website pending - 069
Local Inference Residency
An architectural strategy restricting model inference computation entirely within data centers in a specific geographic region t
Website pending - 070
Machine-Readable Government
A governance model converting policies into structured data formats, enabling AI systems to automate compliance assessments whil
Website pending - 071
Agent Engineering Framework
A system architecture for building, orchestrating, and managing autonomous agents, coordinating model reasoning, tool invocation
Website pending - 072
Cross-modal Representation Alignment
A metric measuring the consistency of vector space distributions across different modality models when processing identical sema
Website pending - 073
PFLOPS
A metric for measuring floating-point operations per second (quadrillions), commonly used to evaluate the compute scale of super
Website pending - 074
Visual Perception Robustness
The ability of computer vision systems to maintain accurate recognition despite disturbances such as lighting changes, occlusion
Website pending - 075
L4 Autonomous Driving
A level of automation where the system performs all driving tasks within a specific Operational Design Domain without human inte
Website pending - 076
HVDC Power Supply
An architecture delivering high-voltage DC directly to loads, reducing AC-DC conversion losses and improving data center energy
Website pending - 077
Probabilistic Constellation Shaping 16-QAM
A 16-QAM modulation format combined with Probabilistic Constellation Shaping to enhance spectral efficiency and transmission rea
Website pending - 078
Hollow-Core Fiber
A fiber structure where light propagates primarily through air holes, offering lower latency and higher nonlinear thresholds tha
Website pending - 079
CUDA Library Invocation
Directly linking and using NVIDIA's pre-compiled high-performance math libraries (e.g., cuBLAS) within high-level frameworks to
Website pending - 080
JIT Compilation
A technique that dynamically compiles intermediate code into native machine code at runtime, aiming to balance startup speed and
Website pending