Wiki
Understand AI concepts and follow the links to practical tools and examples.
103个词条
Showing 103 of 103 entries
Clear search ×AI Concepts
18-
Agent Memory
Agent Memory · Long-term Memory
Agent memory stores and retrieves information such as task progress, preferences, or verif…
-
Constitutional AI (CAI)
Constitutional AI · CAI
Constitutional AI (CAI) is a training paradigm Anthropic proposed in 2022, aiming to repla…
-
Context Engineering
Context Engineering · Context Management
Context engineering selects, organizes, and maintains information for model calls: instruc…
-
In-Context Learning (ICL)
In-Context Learning · ICL
In-context learning (ICL) is one of the most important capabilities of LLMs: putting a few…
-
Mechanistic Interpretability
Mechanistic Interpretability · Interpretability
Mechanistic interpretability (mech interp) is a branch of AI safety and alignment research…
-
Mixture of Experts (MoE)
Mixture of Experts · MoE
Mixture of Experts (MoE) is a neural network design based on conditional computation and s…
-
Model Calibration
Calibration · Model Calibration
Calibration measures whether a model's self-reported confidence actually matches its accur…
-
Model Context Protocol (MCP)
Model Context Protocol · MCP
The Model Context Protocol (MCP) is an open-standard communication protocol open-sourced b…
-
Prompt Injection
Prompt Injection · Indirect Prompt Injection
Prompt Injection is the primary security vulnerability afflicting Large Language Model ( L…
-
RLVR (Reinforcement Learning with Verifiable Rewards)
RLVR · Reinforcement Learning with Verifiable Rewards
RLVR is the new RL paradigm for large models that emerged from 2024 onward: the reward sig…
-
Self-Play
Self-Play · Self Play
Self-play is a reinforcement-learning training paradigm: agents play against / collaborate…
-
Speculative Decoding
Speculative Decoding · speculative sampling
Speculative decoding (also known as speculative sampling) is an inference acceleration tec…
-
Structured Output
Structured Output · Structured Outputs
Structured output returns data in an agreed shape for programmatic consumption. It address…
-
Synthetic Data
Synthetic Data · Data Synthesis
Synthetic data is training data created by a model (usually a stronger model or a speciali…
-
System 1 / System 2 (Dual-Process Thinking)
System 1 · System 2
System 1 / System 2 is the cognitive framework psychologist Daniel Kahneman laid out in Th…
-
Tool Use
Tool Use · Tool Calling
Tool use lets a model request operations such as search, file access, and database queries…
-
Transformer Architecture
transformer · self-attention model
The Transformer is a landmark neural network architecture introduced by Google in the 2017…
-
World Model
World Model · World Foundation Model
A world model is a visionary concept in AI research: building an internal representation o…
Concepts
38-
AGI (Artificial General Intelligence)
AGI · general AI
AGI (Artificial General Intelligence) refers to artificial intelligence that matches or ex…
-
AI Agent
Agent · autonomous agent
An AI Agent (or Autonomous Agent) is an intelligent software entity that employs a Large L…
-
AI Alignment
alignment
AI Alignment is the field of research that makes AI systems' behavior consistent with huma…
-
AI Search (Answer Engine)
answer engine · AI search
AI Search (Answer Engine) refers to search products that use an LLM to directly generate a…
-
BLEU / ROUGE
BLEU · ROUGE
BLEU and ROUGE are the most classic automatic metrics for NLP text generation tasks. Still…
-
BM25
BM25 · Okapi BM25
BM25 (Best Matching 25, or Okapi BM25) is a ranking function originating from the Probabil…
-
CUDA
CUDA · GPU programming
CUDA is NVIDIA's GPU parallel computing API. Deep learning depends almost entirely on CUDA…
-
Chain-of-Thought
CoT · reasoning chain
Chain-of-Thought (CoT) is the prompting technique that makes an LLM explicitly output inte…
-
Code Generation
code-gen · AI coding
Code Generation is the capability of an LLM to write runnable code from natural language d…
-
Context Window
context length · context window
The context window (or context length) defines the maximum continuous span of tokens ($L_{…
-
Dataset
training data · dataset card
A dataset is a collection of examples organized for a task, such as text, images, labels, …
-
Document Chunking
Chunking · Text Splitting
Document Chunking represents the foundational data engineering pipeline in Retrieval-Augme…
-
Embedding
embeddings · dense vector
An embedding is a mathematical technique that maps discrete, high-dimensional symbols (suc…
-
Emergent Abilities
emergence · emergent capabilities
Emergent abilities describe cases where task performance appears to change from weak or ne…
-
Fine-tuning
fine-tuning · fine-tune
Fine-tuning is a transfer learning process that continues updating the neural network weig…
-
FlashAttention
Flash Attention · FlashAttention-2
FlashAttention is an exact, IO-aware attention algorithm engineered around GPU memory hier…
-
Function Calling
function calling · tool use
Function Calling (also termed Tool Use or Tool Calling) is the foundational mechanism that…
-
Hallucination
hallucination · confabulation
In Large Language Models ( LLMs ), hallucination refers to generating content that appears…
-
Large Language Model (LLM)
LLM · large language model
A Large Language Model (LLM) is an autoregressive deep neural network—typically ranging fr…
-
Mamba (State Space Model)
state space model · SSM
Mamba is a deep sequence modeling architecture founded on Selective State Space Models (Se…
-
Model Distillation
knowledge distillation
Model Distillation is the process of "distilling" knowledge from a large model (teacher) i…
-
Model Inference
LLM serving · model inference
Model inference (in enterprise production environments termed LLM Serving) is the systems …
-
Multimodal Models
MLLM · VLM
Multimodal models process multiple data types simultaneously: text, images, audio, video. …
-
Open-Source and Open-Weight LLMs
open-source LLM · open weights
Open-source and open-weight language models are often described interchangeably. Downloada…
-
PPO (Proximal Policy Optimization)
PPO · Proximal Policy Optimization
PPO (Proximal Policy Optimization) is the reinforcement learning algorithm proposed by Ope…
-
Perplexity
PPL · perplexity
Perplexity (PPL) is the most fundamental metric for evaluating language models. It measure…
-
Pre-training
Pretraining · pre-training
Pre-training represents the most compute-intensive and architecturally intricate phase in …
-
Prompt Engineering
prompt design
Prompt Engineering is the practice of designing the input to an LLM to get desired outputs…
-
Reasoning Model
Reasoning Model · slow thinking model
A reasoning model is an architectural specialization within the Large Language Model ( LLM…
-
Reinforcement Learning from Human Feedback (RLHF)
RLHF · Reinforcement Learning from Human Feedback
Reinforcement Learning from Human Feedback (RLHF) is an alignment methodology that trains …
-
Reranking
rerank · cross-encoder
Reranking (or Rerank) re-evaluates and re-orders an initial candidate list retrieved by an…
-
Retrieval-Augmented Generation (RAG)
RAG · retrieval-augmented generation
Retrieval-Augmented Generation (RAG) is a system architecture that tightly couples externa…
-
SWE-bench
SWE-bench Verified
SWE-bench (Software Engineering Benchmark) is the most authoritative code engineering benc…
-
Scaling Laws
scaling law · Chinchilla
Scaling laws describe empirical power-law relationships governing how test cross-entropy l…
-
Self-Attention
attention · Self-Attention
Self-Attention is the core computational unit of the Transformer architecture . It lets ev…
-
Supervised Fine-Tuning (SFT)
SFT · Supervised Fine-Tuning
Supervised Fine-Tuning (SFT) , also designated as Instruction Tuning , represents the pivo…
-
Token
tokens · word piece
A token is the smallest unit of text that a large language model processes. One token roug…
-
Tokenization
tokenizer · BPE
Tokenization is the foundational stage of any modern large language model ( LLM ) data pip…
Infrastructure
11-
Docker
Docker · container
Docker is the de facto standard containerization platform. Packages applications
-
Elasticsearch
ES · Elastic
Elasticsearch (ES) is the distributed search engine built on Apache Lucene, the most popul…
-
Hugging Face
HuggingFace · HF
Hugging Face (huggingface.co) is the "GitHub + AWS" of AI: open-source model hosting, data…
-
LangChain
LangGraph · LangSmith
LangChain is the most popular LLM application development framework, covering the full pip…
-
LlamaIndex
GPT Index
LlamaIndex is a framework focused on connecting LLMs with private data, the mainstream cho…
-
Observability
Observability · Tracing
Observability uses runtime information to understand system behavior. For LLM applications…
-
Ollama
Ollama
Ollama is the simplest tool for running open-source LLMs locally or at the edge. One comma…
-
Qdrant
vector search engine · Qdrant database
Qdrant is an open-source vector database and search engine implemented in Rust. It stores …
-
Vector Database
vector DB · vector retrieval
A vector database is specialized data infrastructure designed to store, index, and query h…
-
pgvector (PostgreSQL Vector Extension)
pgvector · PostgreSQL Vector Search
pgvector is the definitive open-source vector similarity search extension for PostgreSQL. …
-
vLLM
vLLM · PagedAttention
vLLM is an open-source, high-throughput, low-latency LLM serving and inference engine deve…
Models & Products
24-
ZICQ (智客)
ZICQ · Zhike
ZICQ (智客, domain: zicq.com) is a bilingual (Chinese & English) artificial intelligence tec…
-
Anthropic
Anthropic · Anthropic AI
Anthropic is a frontier AI safety company founded in 2021 by former OpenAI VP of Research …
-
ChatGPT
ChatGPT · chat.openai.com
ChatGPT is OpenAI's conversational AI product launched in November 2022, the milestone tha…
-
Claude
Anthropic · Claude 3.5
Claude is Anthropic's large language model family, known for safety, reasoning ability, an…
-
Cohere
Cohere · Cohere Command
Cohere is a Canadian AI company (founded 2020), focused on enterprise-grade LLMs and RAG t…
-
Cursor
Cursor · Cursor IDE
Cursor is the AI-first code editor from Anysphere, based on a VS Code fork. In 2024-2025 i…
-
DeepSeek
DeepSeek · DeepSeek V3
DeepSeek is a foundation AI research lab and tech company based in Hangzhou, China. Betwee…
-
Fireworks AI
Fireworks AI · Fireworks
Fireworks AI is a US-based frontier-model inference platform founded in 2022, focused on s…
-
GPT-5
OpenAI GPT-5
GPT-5 is OpenAI's flagship large language model released in 2025. Compared to GPT-4, it of…
-
Gemini
Google Gemini · Gemini 1.5
Gemini is Google DeepMind's large language model family released at end of 2023, Google's …
-
Google DeepMind
DeepMind · Google DeepMind
Google DeepMind is Google's AI research lab, descended from DeepMind Technologies, founded…
-
Groq
Groq · Groq LPU
Groq is an American AI inference chip + cloud service company. Its self-developed LPU (Lan…
-
Llama
LLaMA · Meta Llama
Llama is Meta's open-source large language model family, currently the benchmark for open-…
-
Meta AI
Meta AI · FAIR
Meta AI is the AI division of Meta (Facebook's parent company), tracing back to FAIR (Face…
-
MiniMax
MiniMax · Hailuo
MiniMax (Shanghai Xiyu Technology / 稀宇科技) is a leading Chinese multimodal foundation-model…
-
Mistral AI
Mistral · Mistral 7B
Mistral AI is a Paris, France AI company (founded 2023), focused on open-source LLMs and i…
-
NVIDIA
NVIDIA · Jensen Huang
NVIDIA is the GPU and accelerated-computing company founded in 1993 by Jensen Huang (黄仁勋).…
-
OpenRouter
OpenRouter · model router
OpenRouter (openrouter.ai) is an AI model API aggregation platform. It unifies models from…
-
Qwen (Tongyi Qianwen)
Qwen · Qwen2.5
Qwen (通义千问) is Alibaba Cloud's LLM family, with Chinese capability among the strongest of …
-
Stability AI
Stability AI · Stable Diffusion
Stability AI is a London-based open-source generative AI company founded in 2019 by Emad M…
-
Together AI
Together AI · Together
Together AI is a San Francisco-based frontier-model cloud platform founded in 2022 with th…
-
TypeSafe AI
TypeSafe AI · Jev
TypeSafe AI is an AI lab that came out of stealth on 2026-09-15, founded by former OpenAI …
-
Whisper
Whisper · OpenAI Whisper
Whisper is OpenAI's open-source automatic speech recognition (ASR) model released in 2022.…
-
xAI
xAI · Grok
xAI is the AI lab founded by Elon Musk in 2023 with the stated mission of "understanding t…
Security
3-
Data Poisoning
data poisoning · training data attack
Data Poisoning means attackers inject malicious samples into pretraining / SFT / RLHF data…
-
Jailbreak
jailbreak · prompt jailbreak
Jailbreak means using carefully crafted prompts to bypass LLM safety alignment, making the…
-
Red Team
red team · adversarial eval
Red Team means organizing specialized teams (internal or external) to simulate attackers, …
Technical Terms
6-
Benchmark
benchmark · evaluation
A benchmark specifies tasks, data, and scoring rules. A score describes performance under …
-
Direct Preference Optimization (DPO)
Direct Preference Optimization · DPO
Direct Preference Optimization (DPO) is a post-training alignment algorithm introduced by …
-
JSON Mode
JSON Mode · JSON Output Mode
JSON mode is an interface feature intended to produce valid JSON. It differs from structur…
-
Key-Value Cache (KV Cache)
Key-Value Cache · KV Cache
The Key-Value Cache (KV Cache) is an indispensable memory-centric acceleration mechanism i…
-
Low-Rank Adaptation (LoRA)
Low-Rank Adaptation · LoRA
LoRA (Low-Rank Adaptation) is a foundational Parameter-Efficient Fine-Tuning (PEFT) method…
-
Model Quantization
Quantization · GPTQ
Model Quantization is an essential deep learning systems optimization technique that reduc…
Terminology
3-
Data Cleaning
deduplication · data preprocessing
Data cleaning identifies and handles duplicates, malformed records, missing values, and co…
-
Data Leakage and Evaluation Contamination
benchmark contamination · train-test leakage
Data leakage occurs when training, preprocessing, or model selection uses information that…
-
Hybrid Search
hybrid retrieval · RRF
Hybrid search combines results from multiple retrieval methods. A common setup pairs lexic…
没有匹配的词条 — 换个关键词试试。Browse all entries