Research map / Multi-agent systems and games / Human-AI and goal-conditioned
QuantV2X: A Fully Quantized Multi-Agent System for Cooperative Perception
Seth Zhao, Huizhi Zhang, Zhaowei Li, Juntong Peng, Anthony Chui, Zewei Zhou, Zonglin Zonglin, Hao Xiang, Zhiyu Huang, Fujia Wang, Ran Tian, Chenfeng Xu and 2 more
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 |
|---|---|---|---|
| quantize model and transmit codebook indices instead of FP32 BEV features | Pyramid Fusion · V2X-Real system-level | mAP30: 43.1 -> 52.6 | Improved |
| model quantization FP32 to INT8 | Pyramid Fusion · V2X-Real system-level | T_comp latency: 59.5ms -> 27.1ms | Improved |
| model quantization FP32 to INT8 (also shrinks payload) | Pyramid Fusion · V2X-Real system-level | mAP30: 43.1 -> 48.7 | Improved |
| add codebook-index communication on top of quantization | Pyramid Fusion · V2X-Real system-level | T_comm / total latency: 286.8ms -> 12.9ms; 346.3ms -> 40.0ms | Improved |
| add AdaRound to max-min calibration (ablation) | Pyramid Fusion · DAIR-V2X (LP+LS) INT4/8 | AP30: 76.5 -> 80.1 | Improved |
| add AdaRound to max-min calibration (ablation) | Pyramid Fusion · DAIR-V2X (LP+CR) INT4/8 | AP30: 73.2 -> 72.8 | Hurt |
| add heterogeneity alignment loss (ablation) | Pyramid Fusion · DAIR-V2X (LP+CR) INT4/8 | AP30: 72.8 -> 74.0 | Improved |
| add spatial alignment loss (ablation) | Pyramid Fusion · DAIR-V2X (LP+CR) INT4/8 | AP30: 74.0 -> 74.2 | Improved |
| add alignment module (hetero and spatial losses) (ablation) | Pyramid Fusion · DAIR-V2X (LP+CR) long range AP70 50m+ | AP70: 22.0 -> 27.9 | Improved |
| add spatial alignment loss (ablation) | Pyramid Fusion · DAIR-V2X (LP+LS) INT4/8 | AP30: 80.8 -> 80.2 | No clear effect |
| quantize V2X-ViT to INT8 weights and activations | V2X-ViT · DAIR-V2X (LP+CR) | AP30/AP50: 57.4/49.5 -> 40.0/11.0 | Hurt |
| lower weight precision from INT8 to INT2 | Pyramid Fusion · DAIR-V2X (LP+CR) | AP30/AP50: 74.6 / 67.8 -> 40.8 / 37.0 | Hurt |
Closest papers by meaning
- An Extensible Framework for Open Heterogeneous Collaborative Perception (ICLR 2024 · similarity 0.88)
- CoSDH: Communication-Efficient Collaborative Perception via Supply-Demand Awareness and Intermediate-Late Hybridization (CVPR 2025 · similarity 0.87)
- STAMP: Scalable Task- And Model-agnostic Collaborative Perception (ICLR 2025 · similarity 0.87)
- CoPEFT: Fast Adaptation Framework for Multi-Agent Collaborative Perception with Parameter-Efficient Fine-Tuning (AAAI 2025 · similarity 0.86)
- Privacy-Preserving V2X Collaborative Perception Integrating Unknown Collaborators (AAAI 2025 · similarity 0.86)
- Pragmatic Heterogeneous Collaborative Perception via Generative Communication Mechanism (NeurIPS 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.
