« Apprentissage par renforcement verbal » : différence entre les versions


(Page créée avec « ==en construction== == Définition == XXXXXXXXX == Français == ''' XXXXXXXXX ''' == Anglais == ''' Verbal Reinforcement Learning''' Large language models (LLMs) have been increasingly used to interact with external environments (e.g., games, compilers, APIs) as goal-driven agents. However, it remains challenging for these language agents to quickly and efficiently learn from trial-and-error as traditional reinforcement learning methods require extensive tra... »)
 
Aucun résumé des modifications
Ligne 2 : Ligne 2 :


== Définition ==
== Définition ==
XXXXXXXXX
L'apprentissage par renforcement verbal est un type d''''[[apprentissage par renforcement]]''' qui donne un retour d'information verbale ou linguistique qui peut être à la fois plus nuancé et spécifique que les récompenses scalaires utilisées dans l'apprentissage par renforcement traditionnel. Ce type d'apprentissage est principalement utilisé pour renforcer les agents linguistiques ainsi que pour augmenter leur capacité.


== Français ==
== Français ==
''' XXXXXXXXX '''
''' apprentissage par renforcement verbal '''


== Anglais ==
== Anglais ==
''' Verbal Reinforcement Learning'''
''' verbal reinforcement learning'''


Large language models (LLMs) have been increasingly used to interact with external environments (e.g., games, compilers, APIs) as goal-driven agents. However, it remains challenging for these language agents to quickly and efficiently learn from trial-and-error as traditional reinforcement learning methods require extensive training samples and expensive model fine-tuning. We propose Reflexion, a novel framework to reinforce language agents not by updating weights, but instead through linguistic feedback. Concretely, Reflexion agents verbally reflect on task feedback signals, then maintain their own reflective text in an episodic memory buffer to induce better decision-making in subsequent trials. Reflexion is flexible enough to incorporate various types (scalar values or free-form language) and sources (external or internally simulated) of feedback signals, and obtains significant improvements over a baseline agent across diverse tasks (sequential decision-making, coding, language reasoning). For example, Reflexion achieves a 91% pass@1 accuracy on the HumanEval coding benchmark, surpassing the previous state-of-the-art GPT-4 that achieves 80%. We also conduct ablation and analysis studies using different feedback signals, feedback incorporation methods, and agent types, and provide insights into how they affect performance.  
''Verbal reinforcement learning is a type of reinforcement learning that provides verbal or linguistic feedback that can be both more nuanced and specific than the scalar rewards used in traditional reinforcement learning. This type of learning is mainly used to reinforce linguistic agents and increase their capacity.''




Ligne 17 : Ligne 17 :
[https://arxiv.org/abs/2303.11366  Source : arxiv]
[https://arxiv.org/abs/2303.11366  Source : arxiv]


[https://medium.com/@vikram40441/implementing-reflexion-language-agents-with-verbal-reinforcement-learning-e4cb300278b6  Source : Medium]




[[Catégorie:vocabulary]]
 
[[Catégorie:publication]]

Version du 30 septembre 2024 à 10:48

en construction

Définition

L'apprentissage par renforcement verbal est un type d'apprentissage par renforcement qui donne un retour d'information verbale ou linguistique qui peut être à la fois plus nuancé et spécifique que les récompenses scalaires utilisées dans l'apprentissage par renforcement traditionnel. Ce type d'apprentissage est principalement utilisé pour renforcer les agents linguistiques ainsi que pour augmenter leur capacité.

Français

apprentissage par renforcement verbal

Anglais

verbal reinforcement learning

Verbal reinforcement learning is a type of reinforcement learning that provides verbal or linguistic feedback that can be both more nuanced and specific than the scalar rewards used in traditional reinforcement learning. This type of learning is mainly used to reinforce linguistic agents and increase their capacity.


Source

Source : arxiv

Source : Medium

Contributeurs: Arianne , wiki