AI Factory Glossary

knowledge graph,kg,entity,relation

Knowledge graphs store entities and relations. Query for structured facts to augment LLM with precise knowledge.

knowledge graph,ontology,semantic

Knowledge graphs represent facts as nodes and edges. Enable reasoning. Combine with LLMs for grounding.

knowledge localization, explainable ai

Find where knowledge is stored.

knowledge masking, nlp

Mask factual knowledge during training.

knowledge neuron, interpretability

Knowledge neurons are individual units encoding specific factual knowledge identifiable through causal intervention.

knowledge neurons, explainable ai

Neurons storing factual knowledge.

knowledge retention techniques, continual learning

Preserve learned information.

known good die (kgd),known good die,kgd,testing

Dies tested and verified before packaging.

known good die for chiplets, kgd, advanced packaging

Pre-tested dies for multi-die packages.

koala,berkeley,dialog

Koala is Berkeley dialog model. Web-scraped conversations.

koboldcpp,roleplay,local

KoboldCpp runs LLMs for roleplay and writing. Built on llama.cpp.

koh etch,etch

Potassium hydroxide for anisotropic silicon etching along crystal planes.

kolmogorov-arnold networks (kan),kolmogorov-arnold networks,kan,neural architecture

Networks using learnable activation functions on edges.

koopman operator theory, control theory

Linear representation of nonlinear dynamics.

kosmos,multimodal ai

Multimodal large language model from Microsoft.

krf (krypton fluoride),krf,krypton fluoride,lithography

248nm DUV light source.

krippendorff's alpha, evaluation

Krippendorff's alpha measures agreement accounting for chance and data types.

krippendorff's alpha,evaluation

Reliability coefficient for multiple annotators.

krum aggregation, federated learning

Select closest updates.

kruskal-wallis, quality & reliability

Kruskal-Wallis test compares distributions across multiple groups.

kserve,kubernetes,inference

KServe is Kubernetes inference platform. Serverless, autoscaling.

kubeflow, infrastructure

Kubernetes-based ML workflow platform.

kubeflow,kubernetes,ml

Kubeflow runs ML workflows on Kubernetes. Pipelines, training, serving. Google origin.

kubernetes for ml,infrastructure

Use K8s to deploy and manage ML workloads.

kv cache eviction, kv, optimization

Remove old entries from cache.

kv cache management,optimization

Efficiently store and reuse key-value pairs.

kv cache optimization, kv, optimization

Efficiently manage key-value cache.

kv cache quantization, kv, optimization

Reduce KV cache memory.

kv cache, kv, llm optimization

KV cache stores key-value attention states avoiding recomputation during generation.

kv cache,key value,memory

KV cache stores computed keys and values for past tokens. Grows with sequence length. Major memory consumer.

kv cache,llm architecture

Cached key-value pairs from previous tokens to speed up autoregressive generation.

kv cache,prefix caching,cache

KV cache stores computed keys/values. Prefix caching reuses cache for shared prompts. Reduces redundant computation.