The dependency graph

Modern AI is a network, not a list. Start with the executive view — the ~15 works that form the spine — then drill into a domain, a single paper's neighbourhood, or the full graph. Node colour marks the research domain; a dark ring marks a reviewed record. Every relationship is also available as a structured table below.

Strong evidence only15 nodes · 9 edges

Loading graph…

View as a structured list (9 relationships)

Dependency relationships shown in the graph: source work, relationship type, target work, and evidence tier.
From	Relationship	To	Evidence
Attention Is All You Need	applies to	BERT	direct
Attention Is All You Need	applies to	Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer	direct
Language Models are Few-Shot Learners	provides evidence for	Scaling Laws for Neural Language Models	direct
Scaling Laws for Neural Language Models	challenges	Training Compute-Optimal Large Language Models	direct
InstructGPT	replaces	Direct Preference Optimization	direct
Language Models are Few-Shot Learners	enables	Chain-of-Thought Prompting Elicits Reasoning	direct
Chain-of-Thought Prompting Elicits Reasoning	combines	ReAct	direct
Training Compute-Optimal Large Language Models	enables	LLaMA	direct
Attention Is All You Need	makes efficient	FlashAttention	direct