Research map / Label-efficient and robust learning / Recognition and long-tail
Learning Personalized End-to-End Goal-Oriented Dialog
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 profile embedding to the MemN2N (ablation) | MemN2N · Personalized bAbI Task 3: Displaying Options | per-response accuracy: 58.94 -> 71.00 | Helped |
| Add similar-user global memory to the MemN2N (ablation) | MemN2N · Personalized bAbI Task 3: Displaying Options | per-response accuracy: 58.94 -> 71.01 | Helped |
| Add profile embedding and global memory (ablation) | MemN2N · Personalized bAbI Task 3: Displaying Options | per-response accuracy: 58.94 -> 71.12 | Helped |
| Add a profile-derived preference bias for knowledge-base entities (ablation) | MemN2N · Personalized bAbI Task 4: Providing Information | per-response accuracy: 57.17 -> 81.38 | Helped |
| Combine the profile model and preference model (ablation) | Personalized MemN2N · Personalized bAbI Task 5: Full Dialog | per-response accuracy: 87.28 -> 95.33 | Helped |
| Use global memory from similar users rather than random users (ablation) | Global Memory · Personalized bAbI Task 5: Full Dialog | per-response accuracy: 87.17 -> 91.70 | Helped |
| Use profile embedding without global memory rather than the combined method (ablation) | Personalized MemN2N · Personalized bAbI Task 5: Full Dialog | per-response accuracy: 95.33 -> 93.83 | Hurt |
| Use global memory without profile embedding rather than the combined method (ablation) | Personalized MemN2N · Personalized bAbI Task 5: Full Dialog | per-response accuracy: 95.33 -> 91.70 | Hurt |
| Use the profile model without the preference model (ablation) | Personalized MemN2N · Personalized bAbI Task 4: Providing Information | per-response accuracy: 81.56 -> 57.18 | Hurt |
| Use the preference model without the profile model (ablation) | Personalized MemN2N · Personalized bAbI Task 4: Providing Information | per-response accuracy: 81.56 -> 81.38 | Hurt |
| Personalize the response system | MemN2N · Human evaluation | task completion rate: 27.6% higher | Helped |
| Personalize the response system | MemN2N · Human evaluation | user satisfaction: 14.3% higher | Helped |
Closest papers by meaning
- Mem-PAL: Towards Memory-based Personalized Dialogue Assistants for Long-term User-Agent Interaction (AAAI 2026 · similarity 0.87)
- Teaching Language Models to Evolve with Users: Dynamic Profile Modeling for Personalized Alignment (NeurIPS 2025 · similarity 0.87)
- Enhancing Personalized Multi-Turn Dialogue with Curiosity Reward (NeurIPS 2025 · similarity 0.87)
- Learning to Clarify: Multi-turn Conversations with Action-Based Contrastive Self-Training (ICLR 2025 · similarity 0.84)
- MMPB: It’s Time for Multi-Modal Personalization (NeurIPS 2025 · similarity 0.84)
- Eliciting Human Preferences with Language Models (ICLR 2025 · 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.