Research map / Large language models / Retrieval and knowledge
Wizard of Wikipedia: Knowledge-Powered Conversational Agents
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 retrieved knowledge conditioning to BoW MemNet (ablation) | BoW MemNet · Full Wizard Test Seen, predicted knowledge | R@1: 56.1 -> 71.3 | Helped |
| Add retrieved knowledge conditioning to Transformer (ablation) | Transformer with Reddit pretraining · Full Wizard Test Seen, predicted knowledge | R@1: 79.0 -> 86.8 | Helped |
| Add Reddit pretraining (ablation) | Transformer knowledge selector · Knowledge Selection Unseen Test | R@1: 12.2 -> 23.7 | Helped |
| Add SQuAD training after Reddit pretraining (ablation) | Transformer knowledge selector · Knowledge Selection Unseen Test | R@1: 23.7 -> 22.9 | No clear effect |
| Add SQuAD training to the Reddit-pretrained Transformer MemNet (ablation) | Transformer MemNet · Full Wizard Test Seen, predicted knowledge | R@1: 86.8 -> 87.4 | Helped |
| Add auxiliary knowledge-selection supervision (ablation) | E2E Transformer MemNet · Full Wizard Test Seen, gold knowledge | F1: 33.6 -> 35.5 | Helped |
| Add auxiliary knowledge-selection supervision (ablation) | E2E Transformer MemNet · Knowledge Selection Seen Test | R@1: 13.4 -> 21.1 | Helped |
| Add knowledge dropout (ablation) | Two-Stage Transformer MemNet · Full Wizard Test Seen, predicted knowledge | PPL: 54.8 -> 46.5 | Helped |
| Add knowledge dropout (ablation) | Two-Stage Transformer MemNet · Full Wizard Test Unseen, predicted knowledge | F1: 17.4 -> 17.3 | Hurt |
| Use gold rather than predicted knowledge | Generative Transformer Memory Network · Test Seen | F1: 18.9 -> 30.6 | Helped |
| Use knowledge conditioning instead of no knowledge (ablation) | Generative Transformer · Human evaluation Seen | Rating: 2.11 (1.11) -> 2.92 (1.33) | Helped |
| Use knowledge conditioning instead of no knowledge (ablation) | Retrieval Transformer · Human evaluation Seen | Rating: 3.33 (1.30) -> 3.43 (1.10) | No clear effect |
Closest papers by meaning
- Low-Resource Knowledge-Grounded Dialogue Generation (ICLR 2020 · similarity 0.90)
- Knowledge Graph Grounded Goal Planning for Open-Domain Conversation Generation (AAAI 2020 · similarity 0.90)
- $\tau$-Knowledge: Evaluating Conversational Agents over Unstructured Knowledge (ICML 2026 · similarity 0.88)
- \$OneMillion-Bench: How Far are Language Agents from Human Experts? (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)
- Improving Open-Domain Dialogue Response Generation with Multi-Source Multilingual Commonsense Knowledge (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.