Research map / Large language models / Language, safety and interpretability
Neural Response Generation With Dynamic Vocabularies
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 |
|---|---|---|---|
| Replace static decoding vocabulary with a dynamic vocabulary | DVS2S vs. strongest baseline · Baidu Tieba test | BLEU-1: 6.33 -> 9.89 | Helped |
| Replace static decoding vocabulary with a dynamic vocabulary | DVS2S vs. strongest baseline · Baidu Tieba test | Average: 28.92 -> 34.05 | Helped |
| Replace static decoding vocabulary with a dynamic vocabulary | DVS2S vs. strongest baseline · Baidu Tieba test | Distinct-1: 0.161 -> 0.233 | Helped |
| Jointly learn vocabulary construction and response generation rather than training them separately (ablation) | DVS2S vs. S-DVS2S · Baidu Tieba test | BLEU-1: 8.01 -> 9.89 | Helped |
| Jointly learn vocabulary construction and response generation rather than training them separately (ablation) | DVS2S vs. S-DVS2S · Baidu Tieba human evaluation | rating 0: 0.232 -> 0.094 | Helped |
| Jointly learn vocabulary construction and response generation rather than training them separately (ablation) | DVS2S vs. S-DVS2S · Baidu Tieba test | BLEU-2: 2.94 -> 3.51 | No clear effect |
| Jointly learn vocabulary construction and response generation rather than training them separately (ablation) | DVS2S vs. S-DVS2S · Baidu Tieba test | BLEU-3: 0.93 -> 1.33 | No clear effect |
| Select 1k rather than 0 content words (ablation) | DVS2S · Baidu Tieba test | Average: 23.43 -> 34.05 | Helped |
| Select 1k rather than 100 content words (ablation) | DVS2S · Baidu Tieba test | Average: 33.87 -> 34.05 | Helped |
| Select 3k rather than 1k content words (ablation) | DVS2S · Baidu Tieba test | Average: 34.05 -> 29.56 | Hurt |
| Select 5k rather than 1k content words (ablation) | DVS2S · Baidu Tieba test | Average: 34.05 -> 27.63 | Hurt |
| Select 10k rather than 1k content words (ablation) | DVS2S · Baidu Tieba test | Average: 34.05 -> 26.30 | Hurt |
Closest papers by meaning
- Topic Aware Neural Response Generation (AAAI 2017 · similarity 0.89)
- Hierarchical Recurrent Attention Network for Response Generation (AAAI 2018 · similarity 0.87)
- Multiresolution Recurrent Neural Networks: An Application to Dialogue Response Generation (AAAI 2017 · similarity 0.87)
- A Statistical Framework for Ranking LLM-based Chatbots (ICLR 2025 · similarity 0.86)
- The Consensus Game: Language Model Generation via Equilibrium Search (ICLR 2024 · similarity 0.85)
- Reranking Laws for Language Generation: A Communication-Theoretic Perspective (NeurIPS 2024 · 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.