Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Dialogue Learning With Human-in-the-Loop
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 |
|---|---|---|---|
| Combine reward-based imitation with forward prediction | MemN2N trained with Mechanical Turk feedback · Mechanical Turk questions, r=0 | accuracy: 0.333 -> 0.431; 0.358 -> 0.431 | Helped |
| Combine reward-based imitation with forward prediction | MemN2N trained with Mechanical Turk feedback · Mechanical Turk questions, r=1 | accuracy: 0.375 -> 0.441; 0.358 -> 0.441 | Helped |
| Increase binary reward fraction (ablation) | RBI · Mechanical Turk questions | accuracy: 0.333 -> 0.340 -> 0.365 -> 0.375 | Helped |
| Increase binary reward fraction (ablation) | FP · Mechanical Turk questions | accuracy: 0.358; 0.358; 0.358; 0.358 | No clear effect |
| Continue iterative dataset-batch training | RBI · bAbI Task 6 | test accuracy: 0.74 -> 0.87 -> 0.90 -> 0.96 -> 0.96 -> 0.98 | Helped |
| Train FP without balancing or random exploration across dataset-sized batches | FP · bAbI Task 6 | test accuracy: 0.99 -> 0.96 -> 1.00 -> 0.30 -> 1.00 -> 0.29 | Mixed |
| Add balancing across feedback-response clusters (ablation) | FP · bAbI Task 6 | test accuracy: 0.30 -> 0.97 | Helped |
| Add random exploration with ε=0.5 (ablation) | FP · bAbI Task 6 | test accuracy: 0.30 -> 0.99 | Helped |
| Add random exploration (ablation) | RBI · Simulator, bAbI and WikiMovies Task 6 | test accuracy | Helped |
| Add random exploration (ablation) | FP · Simulator, bAbI and WikiMovies Task 6 | test accuracy | Helped |
| Use larger online batch sizes rather than updating after each episode (ablation) | RBI and FP · Simulator, bAbI and WikiMovies | test accuracy | No clear effect |
| Combine RBI with FP (ablation) | MemN2N · bAbI Task 6, dataset-sized batches | test accuracy: 0.29 -> 0.97 | Helped |
Closest papers by meaning
- Enhancing Personalized Multi-Turn Dialogue with Curiosity Reward (NeurIPS 2025 · similarity 0.87)
- MA-RLHF: Reinforcement Learning from Human Feedback with Macro Actions (ICLR 2025 · similarity 0.86)
- Learning to Clarify: Multi-turn Conversations with Action-Based Contrastive Self-Training (ICLR 2025 · similarity 0.85)
- Decoupling Conversational Dynamics in Full-Duplex Spoken Models through Reinforcement Learning (NeurIPS 2026 · similarity 0.84)
- Generalizing Verifiable Instruction Following (NeurIPS 2025 · similarity 0.84)
- Teaching Models to Improve on Tape (AAAI 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.