A LoRA-tuned LLaMA-3-8B with function-calling style instructions reaches 77.1% success and 87.2% inform on MultiWOZ 2.2 using 10% of the training data.
In Proceedings of the 23rd Annual Meeting of the Special Interest Group on Discourse and Dialogue, pages 351–360, Edinburgh, UK
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Spec-TOD: A Specialized Instruction-Tuned LLM Framework for Efficient Task-Oriented Dialogue Systems
A LoRA-tuned LLaMA-3-8B with function-calling style instructions reaches 77.1% success and 87.2% inform on MultiWOZ 2.2 using 10% of the training data.