Research map / Architectures and efficiency / Mixture of experts
CasMoE: A Cascaded Framework for Efficient MoE Inference on Resource-constrained Devices
Chengcheng Wang, Haowen He, Liang Zhao, Xiaoheng Deng, Lixin Duan, Shaohua Wan
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 EAM retrieval in front of EAP in the cascade planner (ablation) | CasMoE · CoLA | throughput: 2.971 -> 3.074 | Improved |
| add EAM retrieval in front of EAP in the cascade planner (ablation) | CasMoE · CoLA | accuracy: 0.71 -> 0.72 | Improved |
| use weighted expert pattern label (router score times frequency) instead of count-based softmax label (ablation) | CasMoE EAP · CoLA | accuracy: 0.62 -> 0.72 | Improved |
| replace on-demand expert loading with CasMoE prompt-level prefetching | Mixtral-8x7B · GLUE | throughput: 65.13% throughput improvement with over 96.6% performance preserved | Improved |
| use CasMoE instead of Mixtral-Offloading | Mixtral-8x7B · GLUE average (8 tasks) | accuracy: 0.64 -> 0.72 | Improved |
| use CasMoE instead of Mixtral-Offloading | Mixtral-8x7B · MMLU | accuracy: 0.626 -> 0.572 | Hurt |
| use CasMoE instead of Mixtral-Offloading | Mixtral-8x7B · WMT16 | SacreBLEU: 27.720 -> 34.153 | Improved |
| use CasMoE instead of full-expert All-in-GPU | Mixtral-8x7B · WMT16 | SacreBLEU: 43.366 -> 34.153 | Hurt |
| use CasMoE instead of All-in-GPU | Mixtral-8x7B · QQP | accuracy: 0.54 -> 0.50 | Hurt |
Closest papers by meaning
- Oracle-MoE: Locality-preserving Routing in the Oracle Space for Memory-constrained Large Language Model Inference (ICML 2025 · similarity 0.92)
- FlexMoE: One-for-All Nested Intra-Expert Pruning for MoE Language Models (NeurIPS 2026 · similarity 0.91)
- LightMoE: Reducing Mixture-of-Experts Redundancy through Expert Replacing (NeurIPS 2026 · similarity 0.91)
- $\textit{Read-ME}$: Refactorizing LLMs as Router-Decoupled Mixture of Experts with System Co-Design (NeurIPS 2024 · similarity 0.90)
- OpenMoE: An Early Effort on Open Mixture-of-Experts Language Models (ICML 2024 · similarity 0.90)
- Toward Efficient Inference for Mixture of Experts (NeurIPS 2024 · similarity 0.90)
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.
