Research map / Large language models / Language, safety and interpretability
Masking Orchestration: Multi-Task Pretraining for Multi-Role Dialogue Representation Learning
Tianyi Wang, Yating Zhang, Xiaozhong Liu, Changlong Sun, Qiong Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Pretrain with multi-task masking rather than random initialization | Our model · CDD JFR | macro F1: 39.61 -> 45.45 | Improved |
| Pretrain with multi-task masking rather than random initialization | Our model · CSD DAR | macro F1: 76.83 -> 78.28 | Improved |
| Pretrain with multi-task masking rather than random initialization | Our model · EMD DAR | macro F1: 52.78 -> 53.69 | Improved |
| Pretrain with multi-task masking rather than random initialization | Our model · CDD CFG | BLEU-4: 2.94 -> 8.59 | Improved |
| Pretrain with multi-task masking rather than random initialization | Our model · EMD DS | BLEU-4: 14.87 -> 23.20 | Improved |
| Add word prediction during pretraining (ablation) | Our model · CDD JFR | macro F1: 41.66 -> 45.45 | Improved |
| Add role prediction during pretraining (ablation) | Our model · CDD JFR | macro F1: 40.78 -> 45.45 | Improved |
| Add sentence generation during pretraining (ablation) | Our model · CDD JFR | macro F1: 43.81 -> 45.45 | Improved |
| Add reference prediction during pretraining (ablation) | Our model · CDD JFR | macro F1: 41.71 -> 45.45 | Improved |
| Add word prediction during pretraining (ablation) | Our model · CDD CFG | BLEU-4: 8.09 -> 8.59 | Improved |
| Add role prediction during pretraining (ablation) | Our model · CDD CFG | BLEU-4: 8.22 -> 8.59 | Improved |
| Add sentence generation during pretraining (ablation) | Our model · CDD CFG | BLEU-4: 5.58 -> 8.59 | Improved |
Closest papers by meaning
- DialogLM: Pre-trained Model for Long Dialogue Understanding and Summarization (AAAI 2022 · similarity 0.88)
- Multi-View Feature Representation for Dialogue Generation with Bidirectional Distillation (AAAI 2021 · similarity 0.87)
- Filling the Gap of Utterance-aware and Speaker-aware Representation for Multi-turn Dialogue (AAAI 2021 · similarity 0.87)
- Structure-Aware Multimodal Sequential Learning for Visual Dialog (AAAI 2024 · similarity 0.85)
- TA&AT: Enhancing Task-Oriented Dialog with Turn-Level Auxiliary Tasks and Action-Tree Based Scheduled Sampling (AAAI 2024 · similarity 0.84)
- Learning to Clarify: Multi-turn Conversations with Action-Based Contrastive Self-Training (ICLR 2025 · similarity 0.84)
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.
