Skip to main content
TechGene
Technology histories
Stories
Technologies
Graph
⌕
Search technologies, stories or articles
中文
Graph of Large language models
Find a local graph
⌕
Local technology graph
Selected technology
Current range
Outside this range
Drag to pan · Scroll to zoom
Relation filters
evolved from
derived from
depends on
enables
applied in
replaces
specializes
cross domain transfer
implementation of
part of
Select a technology
Deep learning
Reinforcement learning from human feedback
Language-model agents
Large language models
Retrieval-augmented generation
Transformer
Visible relations
Loaded · 5 relations
Deep learning
→
Large language models
applied in
The large-scale pretrained language models covered here use deep networks to learn sequence distributions; deep learning provides the training and representation methods.
Reinforcement learning from human feedback
→
Large language models
applied in
The 2022 InstructGPT study demonstrated reinforcement learning from human feedback in language-model post-training; RLHF is not used in every language model or every form of instruction fine-tuning.
Large language models
→
Language-model agents
applied in
Systems such as ReAct use language models to generate task reasoning and actions, then use environmental feedback in subsequent steps; the model is a system component rather than a complete execution environment.
Retrieval-augmented generation
→
Large language models
applied in
RAG combines externally retrieved material with a pretrained language generation model to supplement parametric knowledge; it is not a prerequisite for large language models.
Transformer
→
Large language models
applied in
Large-scale language models such as GPT-3 use the Transformer architecture; this relation represents a verified approach and does not prescribe a single architecture for all language models.