Figure and Table

The <Figure> and <Table> components provide auto-numbered, captioned, and cross-referenceable figures and tables.

Figures

Wrap any content in a <Figure> with an id and caption:

Reference them inline: see and .

Figures can contain any content — images, diagrams, code blocks, or even React components.

Wrap a markdown table in a <Table> with an id and caption:

Architecture	Reasoning	Memory	Planning	Tool Use	Key Idea
ReAct	Chain-of-thought	Short-term	No	Yes	Interleave reasoning and acting steps
Plan-and-Execute	Hierarchical	Long-term	Yes	Yes	Generate full plan, then execute sequentially
Reflexion	Self-critique	Episodic	Yes	Yes	Reflect on failures and retry with memory
LATS	Tree search	Short-term	Yes	Yes	Monte Carlo tree search over action space

Method	HumanEval	MATH	MMLU	GSM8K	ARC-C	Avg
Base LLM	67.0	50.4	86.4	78.2	89.1	74.2
+ Tool Use	85.4	68.2	88.1	83.5	90.3	83.1
+ RAG	72.1	55.8	91.3	80.1	93.7	78.6
+ Tool + RAG	88.2	71.4	92.0	86.3	94.1	86.4

Reference them inline: architectures in , results in .

Use columns to set all columns to "center", "left", or "right":

Model	Params	Top-1 Accuracy	Top-5 Accuracy	Training Data	Epochs
Small	125M	82.1	95.3	ImageNet-1K	90
Base	350M	88.4	97.1	ImageNet-1K	90
Large	1.3B	93.2	99.0	ImageNet-21K	30

Model	Params	Top-1 Accuracy	Top-5 Accuracy	Training Data	Epochs
Small	125M	82.1	95.3	ImageNet-1K	90
Base	350M	88.4	97.1	ImageNet-1K	90
Large	1.3B	93.2	99.0	ImageNet-21K	30

You can also use standard markdown column alignment syntax (:---, :---:, ---:) for per-column control without the columns prop.