Research map / Label-efficient and robust learning / Knowledge distillation
Multi-View Feature Representation for Dialogue Generation with Bidirectional Distillation
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 |
|---|---|---|---|
| use MRBD instead of Seq2Seq+Att | Seq2Seq+Att · DailyDialog and PersonaChat | Dist-1: 4.054 -> 4.762; 0.854 -> 1.745 | Helped |
| use MRBD instead of Seq2Seq+Att | Seq2Seq+Att · DailyDialog and PersonaChat | Dis-1: 0.148 -> 0.136; 0.601 -> 0.300 | Helped |
| use MRBD instead of Seq2Seq+Att | Seq2Seq+Att · DailyDialog and PersonaChat | Average human rating: 3.303 -> 1.363; 2.993 -> 1.250 | Helped |
| train all students on the same training set instead of using subtasks (ablation) | MRBD · DailyDialog test | Dist-1: 4.762 -> 5.317 | Mixed |
| train students on distinct subtasks rather than the same set (ablation) | MRBD · DailyDialog test | Ent-1: 7.265 -> 8.232 | Helped |
| have each student imitate all peers rather than a random subgroup (ablation) | MRBD · DailyDialog test | Dist-1: 4.762 -> 4.692 | Hurt |
| use unidirectional rather than bidirectional distillation (ablation) | MRBD · DailyDialog test | Dist-1: 4.762 -> 3.836 | Hurt |
| use unidirectional rather than bidirectional distillation (ablation) | MRBD · DailyDialog test | Dis-1: 0.136 -> 0.134 | Mixed |
| set subtask overlap to 25% rather than 0% (ablation) | MRBD · DailyDialog test | Dist-2: 30.592 -> 31.585 | Helped |
| increase subtask overlap beyond 25% (ablation) | MRBD · DailyDialog test | Ent-1: 8.351 -> 7.311; 7.265 | Hurt |
| raise imitation probability from 0.5 to 0.8 (ablation) | MRBD · DailyDialog test | Dist-1: 4.762 -> 4.841 | Helped |
| raise imitation probability to 1.0 so all peers are imitated (ablation) | MRBD · DailyDialog test | Dist-1: 4.762 -> 4.692 | Hurt |
Closest papers by meaning
- Heterogeneous-Branch Collaborative Learning for Dialogue Generation (AAAI 2023 · similarity 0.90)
- Zero-Resource Knowledge-Grounded Dialogue Generation (NeurIPS 2020 · similarity 0.89)
- Best of Both Worlds: Transferring Knowledge from Discriminative Learning to a Generative Visual Dialog Model (NeurIPS 2017 · similarity 0.88)
- Structure-Aware Multimodal Sequential Learning for Visual Dialog (AAAI 2024 · similarity 0.87)
- From Retrieval to Generation: A Simple and Unified Generative Model for End-to-End Task-Oriented Dialogue (AAAI 2024 · similarity 0.87)
- V^2Dial: Unification of Video and Visual Dialog via Multimodal Experts (CVPR 2025 · similarity 0.87)
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.