Autor
Divya Sharma
Divya Sharma schreibt das KI-Glossar von ReviewRamble: bisher 219 Begriffe, jeweils verständlich definiert und mit Quellen.
KI-Glossar 219
- Abliteration
- Agent Skills
- Agent-Framework
- Agent-Harness
- Agent-Hooks
- Agent2Agent-Protokoll
- Agenten-Berechtigungen
- Agenten-Gedächtnis
- Agenten-Orchestrierung
- Agenten-Sandbox
- Agenten-Schleife
- Agentic RAG
- Agentische KI
- AIME (Benchmark)
- Anweisungsdatei für Agenten (AGENTS.md)
- API-Schlüssel
- Approximative Nächste-Nachbarn-Suche
- ARC-AGI
- Artificial Analysis Intelligence Index
- Attention-Mechanismus
- Autoregressives Modell
- Basismodell
- Batch-API
- Benchmark-Kontamination
- BM25
- BrowseComp
- Browser-Agent
- C2PA Content Credentials
- Chain-of-Thought
- Chunking
- Coding-Agent
- Computer Use
- Context Engineering
- Context Rot
- Continuous Batching
- ControlNet
- CPU-Offloading
- Cross-Encoder
- Datenexfiltration (KI-Agenten)
- Datenresidenz
- Deep Research (KI)
- Deepfake
- Diffusions-Sprachmodell
- Diffusionsmodell
- Direct Preference Optimization
- Document Parsing
- Elo-Wertung (Modell-Arenen)
- Embedding
- Embedding-Dimension
- Entscheidungsmodell
- EU AI Act
- Few-Shot Prompting
- Fine-Tuning
- Foundation Model
- Frontier-Modell
- Generative Engine Optimization
- GGUF
- Golden Dataset
- GPQA
- GPU
- GPU-Cloud
- GraphRAG
- Großes Sprachmodell
- Grounding
- Guardrails (KI)
- Guidance Scale (CFG)
- Halluzination (KI)
- HNSW
- Human-in-the-Loop
- HumanEval
- Humanity's Last Exam
- Hybride Suche
- Image-to-Image
- Image-to-Video
- Inferenz
- Inferenzanbieter
- Inpainting und Outpainting
- Instruct-Modell
- Jailbreak (LLM)
- KI-Agent
- KI-Alignment
- KI-Assistent
- KI-Avatar
- KI-Benchmark
- KI-Bildhochskalierung
- KI-Code-Editor
- KI-Crawler
- KI-Lieferkettenangriff
- KI-Red-Teaming
- KI-Suchmaschine
- KI-Trainingsdaten
- KI-Wasserzeichen
- Kontextfenster
- Kontextverdichtung
- Kosinus-Ähnlichkeit
- KV-Cache
- Late Interaction (ColBERT)
- Layoutanalyse
- Leaderboard (KI)
- Lip Sync (KI-Video)
- LLM-as-a-Judge
- LLM-Evaluierung
- LLM-Gateway
- LLM-Observability
- LLM-Routing
- llms.txt
- LMArena
- Lokales LLM
- LoRA
- Matryoshka-Embeddings
- MCP-Client
- MCP-Ressourcen und -Prompts
- MCP-Server
- MCP-Transport (stdio, Streamable HTTP)
- Menschliche Bewertung
- Metadatenfilterung
- Mixture of Experts
- MMLU
- Model Card
- Model Context Protocol
- Model Serving
- Modell-Fallback
- Modellgewichte
- Modelllizenz
- Modellparameter
- MTEB
- Multi-Agenten-System
- Multimodales Modell
- No-Code-KI-App-Builder
- NPU
- Numerische Präzision (FP16, BF16, FP8)
- OAuth für KI-Agenten
- ONNX
- Open-Source-KI
- Open-Weight-Modell
- OpenAI-kompatible API
- Optische Zeichenerkennung
- OSWorld
- Output-Tokens
- PagedAttention
- Parametereffizientes Fine-Tuning
- pass@k
- PII-Schwärzung
- Pretraining
- Prompt Caching
- Prompt Engineering
- Prompt Injection
- Quantisierung
- Query Rewriting
- Rate Limit (API)
- ReAct (Reasoning and Acting)
- Reasoning-Modell
- Regressionstests (LLM-Apps)
- Reranking
- Retrieval-Augmented Generation
- Retrieval-Metriken (Recall@k, nDCG, MRR)
- Reward Model
- RLHF
- Safetensors
- Self-Hosting (KI-Modelle)
- Semantic Cache
- Serverless GPU
- Slash-Befehl
- Small Language Model
- SOC 2
- Sparse Retrieval
- Spec-Driven Development
- Speculative Decoding
- Speech-to-Speech-Modell
- Speech-to-Text
- Sprecherdiarisierung
- State Space Model
- Streaming (LLM-Ausgabe)
- Structured Output
- Subagent
- Supervised Fine-Tuning
- SWE-bench
- Synthetische Daten
- System-Prompt
- Tabellenextraktion
- Temperature (Sampling)
- Tensor-Parallelismus
- Terminal-Bench
- Test-Time Compute
- Text-to-3D
- Text-to-Image
- Text-to-Music
- Text-to-Speech
- Text-to-Video
- Time to First Token
- Token (LLM)
- Token-Preise
- Tokenizer
- Tokens pro Sekunde
- Tool Calling
- Tool Poisoning
- Top-p-Sampling
- TPU
- Transformer (Architektur)
- Unified Memory
- Vektordatenbank
- Vektorsuche
- Vibe Coding
- Vision-Language-Modell
- Voice Agent
- Voice Cloning
- VRAM
- Web-Crawling für KI
- Weight-Space Programming
- Wissensdatenbank (KI)
- Wissensdestillation
- Wissensgraph
- Wissensstichtag
- Workflow oder Agent
- Workflow-Automatisierung (KI)
- World Model
- Zero Data Retention
- Zero-Shot
- τ-bench