Research map / Vision-language and multimodal / Cross-modal learning
Omnidirectional Information Gathering for Knowledge Transfer-Based Audio-Visual Navigation
Jinyu Chen, Wenguan Wang, Si Liu, Hongsheng Li, Yi Yang
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 confidence-aware cross-task policy distillation (CCPD) (ablation) | audio-visual navigation baseline · Matterport3D unheard | SPL: 38.6 -> 45.1 | Improved |
| add omnidirectional information gathering (OIG) (ablation) | baseline with CCPD · Matterport3D unheard | SPL: 45.1 -> 50.8 | Improved |
| add OIG to baseline that already has CCPD (ablation) | baseline with CCPD · Matterport3D unheard | SR: 59.5 -> 59.4 | No clear effect |
| use confidence-aware reweighting instead of plain policy distillation (ablation) | CCPD · Matterport3D unheard | SPL: 37.0 -> 45.1 | Improved |
| use plain policy distillation without confidence reweighting (ablation) | audio-visual navigation baseline · Matterport3D unheard | SPL: 38.6 -> 37.0 | Hurt |
| distill from PointGoal teacher instead of imitating oracle shortest path (ablation) | CCPD · Matterport3D unheard | SPL: 42.2 -> 45.1 | Improved |
| use PointGoal agent directly with pseudo-GPS instead of distillation (ablation) | PointGoal agent · Matterport3D unheard | SPL: 45.1 -> 12.3 | Hurt |
| increase top-k distilled steps from 30 to 50 or 100 (ablation) | CCPD · Matterport3D unheard | SPL: 45.1 -> 39.6 (k=50); 45.1 -> 37.3 (k=100) | Hurt |
| reduce top-k distilled steps from 30 to 10 (ablation) | CCPD · Matterport3D unheard | SPL: 45.1 -> 37.1 | Hurt |
| add OIG to the fused ensemble (ablation) | 4-model fusion · Matterport3D unheard | SPL: 50.0 -> 62.6 | Improved |
| increase number of fused models from 1 to 4 (ablation) | ensemble · Matterport3D unheard | SPL: 40.9 -> 50.0 | Improved |
Closest papers by meaning
- Learning to Set Waypoints for Audio-Visual Navigation (ICLR 2021 · similarity 0.89)
- RAO-Nav: Probing Omni-Language Models for Zero-shot Semantic Audio-Visual Navigation (NeurIPS 2026 · similarity 0.89)
- CAVEN: An Embodied Conversational Agent for Efficient Audio-Visual Navigation in Noisy Environments (AAAI 2024 · similarity 0.88)
- Towards Audio-Visual Navigation in Noisy Environments: A Large-Scale Benchmark Dataset and an Architecture Considering Multiple Sound-Sources (AAAI 2025 · similarity 0.86)
- RILA: Reflective and Imaginative Language Agent for Zero-Shot Semantic Audio-Visual Navigation (CVPR 2024 · similarity 0.85)
- Crab: A Unified Audio-Visual Scene Understanding Model with Explicit Cooperation (CVPR 2025 · similarity 0.83)
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.
