Nichol and Schulman: first-order initialization meta-learning

What changed

Repeatedly sampled a task, trained on it and moved a shared parameter initialization toward the resulting weights, learning an initialization for fast adaptation without differentiating through the inner optimization.

Scientific scope

Repeatedly sampled a task, trained on it and moved a shared parameter initialization toward the resulting weights, learning an initialization for fast adaptation without differentiating through the inner optimization. MAML and earlier meta-learning are inherited. Benchmark adaptation does not establish arbitrary-task generalization.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. MAML and earlier meta-learning are inherited. Benchmark adaptation does not establish arbitrary-task generalization.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• Reptile: a Scalable Metalearning Algorithm
• Nichol and Schulman: first-order initialization meta-learning — original project

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction