Research map / Multi-agent systems and games / Cooperative multi-agent RL
EPOC: Efficient Perception via Optimal Communication
Masoumeh Heidari Kapourchali, Bonny Banerjee
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 |
|---|---|---|---|
| communicate with all agents instead of none (ablation) | EPOC head agent · KARD | accuracy (%): 24.2±1 -> 88.1±1 | Improved |
| communicate with all agents instead of none (ablation) | EPOC head agent · UTD-MHAD (Kinect skeleton) | accuracy (%): 18.6±2 -> 73.1±4 | Improved |
| use learned communication policy instead of full communication (ablation) | EPOC · KARD | accuracy (%): 88.1±1 -> 90.2±3 | Improved |
| use learned communication policy instead of full communication (ablation) | EPOC · UTD-MHAD (Kinect skeleton) | accuracy (%): 73.1±4 -> 80.1±4 | Improved |
| use learned communication policy instead of full communication (ablation) | EPOC · UTD-MHAD multimodal (27 actions) | accuracy (%): 84.6 -> 85.11 | Mixed |
| communicate with all agents instead of none (ablation) | EPOC inertia agent · UTD-MHAD multimodal (27 actions) | accuracy (%): 29.2 -> 84.6 | Improved |
| learned policy reduces communication to a fraction of test cases (ablation) | EPOC · KARD | communication rate (%): 63.54% | Improved |
| learned policy reduces communication to a fraction of test cases (ablation) | EPOC · UTD-MHAD (Kinect skeleton) | communication rate (%): 61.32% | Improved |
| online non-myopic decision-making instead of myopic offline Value of Information (ablation) | EPOC · KARD/UTD-MHAD random failure | accuracy | Improved |
| online updating of agents' models instead of non-adaptive (ablation) | EPOC · KARD/UTD-MHAD persistent random messages | number of communications | Improved |
| use ConvNets (60x10^6 params) instead of EPOC (67x10^3 params) | comparison · UTD-MHAD multimodal (27 actions) | accuracy (%): 85.11 -> 86.97 | Improved |
Closest papers by meaning
- M2I2: Learning Efficient Multi-Agent Communication via Masked State Modeling and Intention Inference (AAAI 2026 · similarity 0.86)
- Learning Multi-Agent Communication with Contrastive Learning (ICLR 2024 · similarity 0.85)
- Actions Speak Louder Than Words: Rate-Reward Trade-off in Markov Decision Processes (ICLR 2025 · similarity 0.85)
- Expressive Multi-Agent Communication via Identity-Aware Learning (AAAI 2024 · similarity 0.84)
- Learning to Communicate Through Implicit Communication Channels (ICLR 2025 · similarity 0.84)
- ProAct: A Benchmark and Multimodal Framework for Structure-Aware Proactive Response (ICML 2026 · 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.
