Research map / Label-efficient and robust learning / Semi-supervised learning
GALAXY: A Generative Pre-trained Model for Task-Oriented Dialog with Semi-supervised Learning and Explicit Policy Injection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Pre-train with explicit dialog-act policy learning, consistency regularization, and gated unlabeled data | GALAXY · In-Car | Comb: from 104.95 to 107.45 | Helped |
| Pre-train with explicit dialog-act policy learning, consistency regularization, and gated unlabeled data | GALAXY · MultiWOZ2.0 | Comb: from 105.05 to 110.35 | Helped |
| Pre-train with explicit dialog-act policy learning, consistency regularization, and gated unlabeled data | GALAXY · MultiWOZ2.1 | Comb: from 105.25 to 110.76 | Helped |
| Add the unlabeled-data gating mechanism (ablation) | GALAXY · MultiWOZ2.0 | Comb: 108.11 to 110.35 | Helped |
| Add the dialog-act prediction loss (ablation) | GALAXY · MultiWOZ2.0 | Comb: 103.27 to 110.35 | Helped |
| Add consistency regularization on unlabeled dialogs (ablation) | GALAXY · MultiWOZ2.0 | Comb: 107.27 to 110.35 | Helped |
| Add both dialog-act prediction and consistency regularization losses (ablation) | GALAXY · MultiWOZ2.0 | Comb: 106.79 to 110.35 | Helped |
| Use dialog-policy-aware GALAXY pre-training instead of other conversation-model pre-training | MultiWOZ2.0 · MultiWOZ2.0 | Comb: 106.53 to 110.35 | Helped |
| Use GALAXY rather than existing systems in the reported low-resource settings | MultiWOZ2.0 · 5%, 10%, 20%, and 50% training data | Success | Helped |
Closest papers by meaning
- TA&AT: Enhancing Task-Oriented Dialog with Turn-Level Auxiliary Tasks and Action-Tree Based Scheduled Sampling (AAAI 2024 · similarity 0.89)
- Task-Oriented Dialog Systems That Consider Multiple Appropriate Responses under the Same Context (AAAI 2020 · similarity 0.88)
- Best of Both Worlds: Transferring Knowledge from Discriminative Learning to a Generative Visual Dialog Model (NeurIPS 2017 · similarity 0.87)
- Efficient Dialog Policy Learning by Reasoning with Contextual Knowledge (AAAI 2022 · similarity 0.87)
- Winnie: Task-Oriented Dialog System with Structure-Aware Contrastive Learning and Enhanced Policy Planning (AAAI 2024 · similarity 0.87)
- Structure-Aware Multimodal Sequential Learning for Visual Dialog (AAAI 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.