Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Align before Collaborate: Mitigating Feature Misalignment for Robust Multi-Agent Perception
Dingkang Yang, Dingkang Yang, Ke Li, Dongling Xiao, Zedian Shao, Peng Sun, Liang Song
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 NEAT plugin to five collaborative perception models | F-Cooper, CORE, AttFuse, Where2comm, CoBEVT · V2XSet, OPV2V, V2V4Real, OPV2V Culver City | AP@0.7: average gains 8.85%, 4.10%, 6.50%, 3.81%, 8.78% | Improved |
| add NEAT to F-Cooper | F-Cooper · OPV2V Culver City | AP@0.7: 48.02 -> 53.28 AP@0.7; 68.68 -> 74.91 AP@0.5 | Improved |
| add NEAT to AttFuse on OPV2V | AttFuse · OPV2V | AP@0.7: 59.82 -> 68.19 | Improved |
| use NEAT instead of SyncNet or CoAlign plugins on CoBEVT | CoBEVT · V2XSet | AP@0.7: 60.20 (CoAlign) -> 65.16 | Improved |
| add deformable feature aggregation (FA) to CoBEVT (ablation) | CoBEVT + NEAT · V2XSet | AP@0.7: 54.30 -> 58.24 | Improved |
| add object-related token generation (TG) (ablation) | CoBEVT + NEAT · V2XSet | AP@0.7: 58.24 -> 59.97 | Improved |
| add importance map generation (MG) (ablation) | CoBEVT + NEAT · V2XSet | AP@0.7: 59.97 -> 62.08 | Improved |
| add query proposal selection (PS) (ablation) | CoBEVT + NEAT · V2XSet | AP@0.7: 62.08 -> 63.25 | Improved |
| add region cross-attention reinforcement (RCR) (ablation) | CoBEVT + NEAT · V2XSet | AP@0.7: 63.25 -> 65.16 | Improved |
| add NEAT to vanilla methods under increasing transmission delay | F-Cooper, AttFuse, Where2comm, CoBEVT · OPV2V | AP@0.7 | Improved |
| add NEAT to vanilla methods under increasing localization error | F-Cooper, AttFuse, Where2comm, CoBEVT · V2XSet | AP@0.7 | 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.
- Where2comm — Where2comm: Communication-Efficient Collaborative Perception via Spatial Confidence Maps (NeurIPS 2022) · better on 1 of 1 shared result
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
- RCP-Bench: Benchmarking Robustness for Collaborative Perception Under Diverse Corruptions (CVPR 2025 · similarity 0.89)
- RCDN: Towards Robust Camera-Insensitivity Collaborative Perception via Dynamic Feature-based 3D Neural Modeling (NeurIPS 2024 · similarity 0.88)
- TraF-Align: Trajectory-aware Feature Alignment for Asynchronous Multi-agent Perception (CVPR 2025 · similarity 0.88)
- DSRC: Learning Density-Insensitive and Semantic-Aware Collaborative Representation Against Corruptions (AAAI 2025 · similarity 0.87)
- Generative Map Priors for Collaborative BEV Semantic Segmentation (CVPR 2025 · similarity 0.87)
- Multi-agent Collaborative Perception via Motion-aware Robust Communication Network (CVPR 2024 · similarity 0.87)
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.
