Research map / Large language models / Retrieval and knowledge
Low-Resource Knowledge-Grounded 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 |
|---|---|---|---|
| Use the disentangled pre-trained model with 1/8 of the knowledge-grounded training data instead of the full-data ITDD baseline | Wizard Test Seen · Wizard Test Seen | F1: 16.2 -> 16.3 | Helped |
| Use the disentangled pre-trained model with 1/8 of the knowledge-grounded training data instead of the full-data TMN baseline | Wizard Test Unseen · Wizard Test Unseen | F1: 14.3 -> 16.0 | Helped |
| Use the disentangled pre-trained model with 1/8 of the knowledge-grounded training data instead of the full-data TMN baseline | Wizard Test Unseen · Wizard Test Unseen | BLEU-4: 0.009 -> 0.035 | Helped |
| Use the pre-trained model with 1/8 data rather than the non-pre-trained model (ablation) | Proposed model · Wizard Test Seen and Test Unseen | performance | Helped |
| Fix pre-trained parameters rather than fine-tuning them with limited grounded dialogue data (ablation) | Proposed model · Wizard Test Seen and Test Unseen | performance | Helped |
| Pre-train the language model on ungrounded dialogue rather than fitting it only with grounded dialogue data (ablation) | Proposed model · Wizard Test Seen and Test Unseen | performance | Helped |
| Pre-train the context encoder and context processor on ungrounded dialogue rather than omitting their pre-training (ablation) | Proposed model · Wizard Test Seen and Test Unseen | performance | Helped |
| Pre-train the knowledge encoder on plain text rather than omitting its pre-training (ablation) | Proposed model · Wizard Test Seen and Test Unseen | performance | Helped |
| Include the language model in the decoding manager rather than removing it (ablation) | Proposed model · Wizard Test Seen and Test Unseen | performance | Helped |
| Include the context processor in the decoding manager rather than removing it (ablation) | Proposed model · Wizard Test Seen and Test Unseen | performance | Helped |
| Include the knowledge processor in the decoding manager rather than removing it (ablation) | Proposed model · Wizard Test Seen and Test Unseen | performance | Helped |
| Disentangle the model to pre-train dialogue and document components, rather than applying pre-training to entangled TMN modules (ablation) | Proposed model versus pre-trained TMN · Wizard | performance | Helped |
Closest papers by meaning
- Zero-Resource Knowledge-Grounded Dialogue Generation (NeurIPS 2020 · similarity 0.94)
- KPT: Keyword-Guided Pre-training for Grounded Dialog Generation (AAAI 2023 · similarity 0.91)
- Improving Open-Domain Dialogue Response Generation with Multi-Source Multilingual Commonsense Knowledge (AAAI 2024 · similarity 0.88)
- From Retrieval to Generation: A Simple and Unified Generative Model for End-to-End Task-Oriented Dialogue (AAAI 2024 · similarity 0.87)
- Improving the Robustness of Knowledge-Grounded Dialogue via Contrastive Learning (AAAI 2024 · similarity 0.87)
- $\tau$-Knowledge: Evaluating Conversational Agents over Unstructured Knowledge (ICML 2026 · 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.