Research map / Probabilistic and scientific modeling / Variational and generative sampling
Improving Variational Encoder-Decoders in Dialogue Generation
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 |
|---|---|---|---|
| add scheduled sampling to collaborative VED (ablation) | CO · DailyDialog and Switchboard | NLL: 223.6 -> 213.2; 289.7 -> 273.4 | Helped |
| add scheduled sampling to collaborative VED (ablation) | CO · DailyDialog and Switchboard | PPL: 26.1 -> 23.8; 36.5 -> 31.8 | Helped |
| add scheduled sampling to collaborative VED (ablation) | CO · DailyDialog | Embedding Average: 0.465 -> 0.539 | Helped |
| add scheduled sampling to collaborative VED (ablation) | CO · Switchboard | Embedding Average: 0.377 -> 0.392 | Helped |
| add word dropout to collaborative VED with scheduled sampling (ablation) | CO+SS · DailyDialog and Switchboard | PPL: 25.1 -> 23.8; 34.4 -> 31.8 | Mixed |
| add word dropout to collaborative VED with scheduled sampling (ablation) | CO+SS · DailyDialog | NLL: 218.7 -> 213.2 | Helped |
| add word dropout to collaborative VED with scheduled sampling (ablation) | CO+SS · Switchboard | NLL: 273.4 -> 273.4 | No clear effect |
| train the AE and CVAE parts iteratively instead of jointly (ablation) | CO+SS · DailyDialog and Switchboard | PPL: 28.5 -> 23.8; 39.6 -> 31.8 | Helped |
| train the AE and CVAE parts iteratively instead of jointly (ablation) | CO+SS · DailyDialog and Switchboard | Embedding Average: 0.420 -> 0.539; 0.347 -> 0.392 | Helped |
| replace collaborative VED with BOW loss | dialogue generation · DailyDialog and Switchboard | NLL: 247.3; 321.1 | Hurt |
| use BOW loss for sentence-level topic guidance | dialogue generation · DailyDialog and Switchboard | Embedding Average: 0.475; 0.340 | Helped |
| reserve free bits equally for every latent dimension | VED · DailyDialog and Switchboard | Embedding Average: 0.423|0.336 | Hurt |
Closest papers by meaning
- Variational Memory Encoder-Decoder (NeurIPS 2018 · similarity 0.91)
- A Hierarchical Latent Variable Encoder-Decoder Model for Generating Dialogues (AAAI 2017 · similarity 0.91)
- Towards Diverse, Relevant and Coherent Open-Domain Dialogue Generation via Hybrid Latent Variables (AAAI 2023 · similarity 0.89)
- DialogWAE: Multimodal Response Generation with Conditional Wasserstein Auto-Encoder (ICLR 2019 · similarity 0.88)
- Data Augmentation for Spoken Language Understanding via Joint Variational Generation (AAAI 2019 · similarity 0.88)
- A Comprehensive Study of Decoder-Only LLMs for Text-to-Image Generation (CVPR 2025 · 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.