# Emergent Abilities

> Emergent abilities are capabilities that appear in large language models above a certain scale but are near chance in smaller ones, such as multi-step arithmetic or benefiting from chain-of-thought prompting.

- Identifier: PTL-0184
- Category: Foundations
- Canonical URL: https://protologue.com/t/emergent-abilities/
- Also known as: emergence, emergent capabilities
- Introduced: 2022

## Description

The idea explained why some prompting techniques only work with large models. It is disputed. Later analysis argued that many apparent jumps are artifacts of all-or-nothing metrics such as exact match, and that smoother metrics show gradual improvement with scale. The practical lesson stands either way, which is that a technique that fails on a small model may work on a larger one, and the reverse.

## Related terms

- [Large Language Model](https://protologue.com/t/large-language-model/)
- [Chain-of-Thought Prompting](https://protologue.com/t/chain-of-thought/)
- [In-Context Learning](https://protologue.com/t/in-context-learning/)
- [Benchmark](https://protologue.com/t/benchmark/)

## Sources

- Wei et al. (2022). Emergent Abilities of Large Language Models. https://arxiv.org/abs/2206.07682
- Schaeffer et al. (2023). Are Emergent Abilities of Large Language Models a Mirage?. https://arxiv.org/abs/2304.15004

## Cite this entry

Protologue. (2026). Emergent Abilities. In Protologue: A Taxonomy of Prompting and LLM Techniques (v1.4.0, PTL-0184). https://protologue.com/t/emergent-abilities/

License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)