Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Self-Motivated Communication Agent for Real-World Vision-Dialog Navigation
Yi Zhu, Yue Weng, Fengda Zhu, Xiaodan Liang, Qixiang Ye, Yutong Lu, Jianbin Jiao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 learned whether-to-ask (WeTA) instead of never asking (ablation) | SCoA · CVDN val seen (oracle path) | Goal Progress (m): 4.1 -> 5.93 | Improved |
| add learned WeTA instead of asking at every step (ablation) | SCoA · CVDN val seen (oracle path) | Goal Progress (m): 5.41 -> 5.93 | Improved |
| add learned WeTA instead of asking at every step (ablation) | SCoA · CVDN val unseen (oracle path) | Goal Progress (m): 1.78 -> 1.94 | Improved |
| use entropy-supervised WeTA instead of random asking (ablation) | SCoA · CVDN val seen (oracle path) | Goal Progress (m): 5.03 -> 5.93 | Improved |
| use WeTA instead of IC3Net softmax asking (ablation) | SCoA · CVDN val seen (oracle path) | Goal Progress (m): 4.83 -> 5.93 | Improved |
| use WeTA instead of When2com (ablation) | SCoA · CVDN val unseen (oracle path) | Goal Progress (m): 1.86 -> 1.94 | Improved |
| use WaTA question scoring instead of RMM question generation with oracle stopping (ablation) | SCoA · CVDN val unseen (mixed path, 2,050 instances) | Goal Progress (m): 8.9 -> 11.19 | Improved |
| add WeTA cross-entropy loss (ablation) | SCoA · CVDN val seen (oracle path) | Goal Progress (m): 5.49 -> 5.93 | Improved |
| add WeTA cross-entropy loss (ablation) | SCoA · CVDN val unseen (oracle path) | Goal Progress (m): 1.87 -> 1.94 | Improved |
| add WaTA KL-divergence distillation loss (ablation) | SCoA · CVDN val seen (oracle path) | Goal Progress (m): 5.78 -> 5.93 | Improved |
| add WaTA KL-divergence distillation loss (ablation) | SCoA · CVDN val unseen (oracle path) | Goal Progress (m): 1.86 -> 1.94 | Improved |
| add oracle during testing instead of self-Q&A | SCoA · CVDN val seen (oracle path) | Goal Progress (m): 5.93 -> 6.74 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- When2com — When2com: Multi-Agent Perception via Communication Graph Grouping (CVPR 2020) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Just Ask: An Interactive Learning Framework for Vision and Language Navigation (AAAI 2020 · similarity 0.90)
- Learning Cooperative Visual Dialog Agents With Deep Reinforcement Learning (ICCV 2017 · similarity 0.89)
- CAVEN: An Embodied Conversational Agent for Efficient Audio-Visual Navigation in Noisy Environments (AAAI 2024 · similarity 0.87)
- Collaborative Instance Object Navigation: Leveraging Uncertainty-Awareness to Minimize Human-Agent Dialogues (ICCV 2025 · similarity 0.87)
- General Scene Adaptation for Vision-and-Language Navigation (ICLR 2025 · similarity 0.86)
- Learning to Clarify: Multi-turn Conversations with Action-Based Contrastive Self-Training (ICLR 2025 · similarity 0.86)
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.
