Knowledge Distillation for Relation Extraction
The project "Knowledge Distillation for Relation Extraction" (KD4RE) deals with the development of efficient information extraction methods. Current, very large language models can often only be implemented as an "off-premise" solution due to high technical requirements. KD4RE therefore includes the development of methods that transfer the capacities of very large language models to smaller models and thus, in addition to lower costs, also new "on-premise " solutions.
from: 1 Juni 2022
until: 30 November 2022