« In-context learning » : différence entre les versions
Aller à la navigation
Aller à la recherche
(Page créée avec « [1] The in-context learning (ICL) ability is formally introduced by GPT-3 : assuming that the language model has been provided with a natural language instruction and/or several task demonstrations, it can generate the expected output for the test instances by completing the word sequence of input text, without requiring additional training or gradient update == Références == * [https://arxiv.org/pdf/2303.18223.pdf] A Survey of Large Language Models ») |
Aucun résumé des modifications Balise : Révocation manuelle |
||
| (2 versions intermédiaires par le même utilisateur non affichées) | |||
| Ligne 2 : | Ligne 2 : | ||
[1] The in-context learning (ICL) ability is formally introduced by GPT-3 : assuming that the language model has been provided with a natural language instruction and/or several task demonstrations, it can generate the expected output for the test instances by completing the word sequence of input text, without requiring additional training or gradient update | [1] The in-context learning (ICL) ability is formally introduced by GPT-3 : assuming that the language model has been provided with a natural language instruction and/or several task demonstrations, it can generate the expected output for the test instances by completing the word sequence of input text, without requiring additional training or gradient update | ||
[[File:A comparative illustration of in-context learning (ICL) and chain-of-thought (CoT) prompting.jpg|500px]] | |||
== Références == | == Références == | ||
* [https://arxiv.org/pdf/2303.18223.pdf] A Survey of Large Language Models | * [https://arxiv.org/pdf/2303.18223.pdf] A Survey of Large Language Models | ||
Version actuelle datée du 27 avril 2023 à 16:40
[1] The in-context learning (ICL) ability is formally introduced by GPT-3 : assuming that the language model has been provided with a natural language instruction and/or several task demonstrations, it can generate the expected output for the test instances by completing the word sequence of input text, without requiring additional training or gradient update
_and_chain-of-thought_(CoT)_prompting.jpg){kind=link}
Références
- [1] A Survey of Large Language Models