Meta-RL — learning to learn.

hard

Answer

$Train agent across a distribution of related tasks so it adapts to new tasks quickly.$
$(\mathrm{RL}^{2})$
$(2) MAML for RL: gradient-based meta-learning that produces initialization enabling few-step adaptation.$
$(3) PEARL: context-conditioned policy with latent task variable.$
$Applications: fast robot adaptation, personalized recommenders.$
$Modern LLM in-context learning is analogous.$

Check yourself — multiple choice

Random
Train across task distribution → fast adaptation to new task; $\mathrm{RL}^{2}$ (LSTM context) / MAML (gradient-based) / PEARL (latent task variable); LLM in-context learning is analog
Same as SL
Not real

Meta-RL: fast adaptation across tasks; $\mathrm{RL}^{2}\; / \;\mathrm{MAML}\; / \;\mathrm{PEARL}$ .

#theory

214 interview questions in this topic.