Research map / Large language models / Language, safety and interpretability
Learning from Easy to Complex: Adaptive Multi-Curricula Learning for Neural 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 |
|---|---|---|---|
| Apply adaptive multi-curricula training | SEQ2SEQ, CVAE, Transformer, HRED and DialogWAE · PersonaChat test | BLEU: SEQ2SEQ 0.316 -> 0.352; CVAE 0.290 -> 0.321; Transformer 0.195 -> 0.329; HRED 0.272 -> 0.308; DialogWAE 0.124 -> 0.171 | Helped |
| Apply adaptive multi-curricula training | SEQ2SEQ, CVAE, Transformer, HRED and DialogWAE · DailyDialog test | BLEU: SEQ2SEQ 0.399 -> 0.617; CVAE 0.406 -> 0.691; Transformer 0.412 -> 0.8063; HRED 0.1746 -> 0.3834; DialogWAE 0.0303 -> 0.0 | Helped |
| Apply adaptive multi-curricula training | SEQ2SEQ, CVAE, Transformer, HRED and DialogWAE · OpenSubtitles test | BLEU: SEQ2SEQ 0.140 -> 0.172; CVAE 0.0522 -> 0.0429; Transformer 0.072 -> 0.050; HRED 0.0498 -> 0.0795; DialogWAE 0.0038 -> 0. | Mixed |
| Use five curricula together rather than no curriculum (ablation) | SEQ2SEQ · DailyDialog validation | BLEU: 0.3599 -> 0.7057 | Helped |
| Add a specificity curriculum (ablation) | SEQ2SEQ curriculum learning · DailyDialog validation | BLEU: 0.3599 -> 0.6866 | Helped |
| Add a repetitiveness curriculum (ablation) | SEQ2SEQ curriculum learning · DailyDialog validation | BLEU: 0.3599 -> 0.5994 | Helped |
| Add a query-relatedness curriculum (ablation) | SEQ2SEQ curriculum learning · DailyDialog validation | BLEU: 0.3599 -> 0.5745 | Helped |
| Add a continuity curriculum (ablation) | SEQ2SEQ curriculum learning · DailyDialog validation | BLEU: 0.3599 -> 0.5662 | Helped |
| Add a model confidence curriculum (ablation) | SEQ2SEQ curriculum learning · DailyDialog validation | BLEU: 0.3599 -> 0.5917 | Helped |
| Replace the learned RL curriculum policy with a random policy (ablation) | SEQ2SEQ adaptive multi-curricula learning · DailyDialog | BLEU: 0.617 -> 0.565 | Hurt |
| Train from complex to easy examples (ablation) | SEQ2SEQ adaptive multi-curricula learning · DailyDialog | BLEU: 0.617 -> 0.354 | Hurt |
Closest papers by meaning
- Multi-View Feature Representation for Dialogue Generation with Bidirectional Distillation (AAAI 2021 · similarity 0.88)
- Zero-Resource Knowledge-Grounded Dialogue Generation (NeurIPS 2020 · similarity 0.87)
- Modelling Hierarchical Structure between Dialogue Policy and Natural Language Generator with Option Framework for Task-oriented Dialogue System (ICLR 2021 · similarity 0.87)
- Learning to Clarify: Multi-turn Conversations with Action-Based Contrastive Self-Training (ICLR 2025 · similarity 0.86)
- Decoupling Conversational Dynamics in Full-Duplex Spoken Models through Reinforcement Learning (NeurIPS 2026 · similarity 0.85)
- From Retrieval to Generation: A Simple and Unified Generative Model for End-to-End Task-Oriented Dialogue (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.