Search papers, labs, and topics across Lattice.
University of Regensburg
4
0
8
Forget scaling laws: cross-lingual transfer for ABSA reveals that LLMs benefit most from training on multiple non-target languages, while smaller models thrive on code-switching.
AnnoABSA slashes ABSA annotation time by using LLMs to suggest annotations, improving suggestion accuracy over time by retrieving similar, already-annotated examples.
Forget expensive manual annotation: LLMs can bootstrap lightweight ABSA models to near state-of-the-art performance with just 50 labeled examples.
Achieve top performance in sentiment analysis by having your LLM vote on its own answers.
