Neural Machine Translation Models Can Learn to be Few-shot Learners

no code implementations • 15 Sep 2023 • Raphael Reinauer, Patrick Simianer, Kaden Uhlig, Johannes E. M. Mosig, Joern Wuebker

The emergent ability of Large Language Models to use a small number of examples to learn to perform in novel domains and tasks, also called in-context learning (ICL).

Domain Adaptation In-Context Learning +4

STAR: A Schema-Guided Dialog Dataset for Transfer Learning

1 code implementation • 22 Oct 2020 • Johannes E. M. Mosig, Shikib Mehri, Thomas Kober

We present STAR, a schema-guided task-oriented dialog dataset consisting of 127, 833 utterances and knowledge base queries across 5, 820 task-oriented dialogs in 13 domains that is especially designed to facilitate task and domain transfer learning in task-oriented dialog.

Transfer Learning Zero-shot Generalization

Where is the context? -- A critique of recent dialogue datasets

1 code implementation • 22 Apr 2020 • Johannes E. M. Mosig, Vladimir Vlasov, Alan Nichol

Recent dialogue datasets like MultiWOZ 2. 1 and Taskmaster-1 constitute some of the most challenging tasks for present-day dialogue models and, therefore, are widely used for system evaluation.

Dialogue Transformers

1 code implementation • 1 Oct 2019 • Vladimir Vlasov, Johannes E. M. Mosig, Alan Nichol

We introduce a dialogue policy based on a transformer architecture, where the self-attention mechanism operates over the sequence of dialogue turns.