« Requête enrichie par graphes de connaissances » : différence entre les versions

Version du 24 septembre 2024 à 16:04

Définition

Peaufinage des résultats d'un grand modèle de langues en enrichissant les requêtes avec des connaissances factuelles générées à partir de règles extraites de graphes de connaissances. Ceci permet de générer des résultats plus précis et contextuellement plus pertinents.

Complément

Entrée en cours de travail.

Les règles sont extraites d'un graphe de connaissances à l'aide d'un algorithme classique de parcours en largeur.

Cette technique se rapproche de la génération augmentée d'information applicative (GAIA).

Français

requête enrichie par graphes de connaissances

requête enrichie par chaine(s) de connaissances

requête de résolution par chaine(s) de connaissances

requête par chaine(s) de connaissances

Anglais

Chain-of-Knowledge prompting

Chain-of-Knowledge framework prompting

CoK promptig

Chain-of-Knowledge

CoK

Source

Source : Zhang et al. 2024

[https://cobusgreyling.medium.com/chain-of-knowledge-prompting-0285ac879ede Source : Chain-Of-Knowledge Prompting

- Medium]

@@ Ligne 28 : / Ligne 28 : @@
 '''CoK'''
-<!-- The CHAIN-OF-KNOWLEDGE framework has two main components: dataset construction and model learning. For dataset construction, the authors first mine compositional rules from knowledge graphs. These rules represent patterns of how different facts can be combined to infer new knowledge. They then select knowledge triples from the graph that match these rules. Finally, they use advanced language models to transform the structured knowledge into natural language questions and reasoning steps.
- For model learning, they initially tried simply fine-tuning LLMs on this data. However, this led to "rule overfitting" where models would apply rules even without supporting facts. To address this, they introduced a trial-and-error mechanism. This simulates how humans explore their internal knowledge when reasoning, by having the model try different rules and backtrack if it lacks key facts.
--->
 == Source ==
@@ Ligne 40 : / Ligne 35 : @@
   - Medium]
+<!-- The CHAIN-OF-KNOWLEDGE framework has two main components: dataset construction and model learning. For dataset construction, the authors first mine compositional rules from knowledge graphs. These rules represent patterns of how different facts can be combined to infer new knowledge. They then select knowledge triples from the graph that match these rules. Finally, they use advanced language models to transform the structured knowledge into natural language questions and reasoning steps.
+ For model learning, they initially tried simply fine-tuning LLMs on this data. However, this led to "rule overfitting" where models would apply rules even without supporting facts. To address this, they introduced a trial-and-error mechanism. This simulates how humans explore their internal knowledge when reasoning, by having the model try different rules and backtrack if it lacks key facts.
+-->
 [[Catégorie:Publication]]

« Requête enrichie par graphes de connaissances » : différence entre les versions