Direct GPU Energy Measurements of Batch-Level Sharing in Mixture-of-Experts and Recurrent Models
cs.LG, cs.AI
Submitted: 2026-11-08
Updated: 2026-11-08
Terminology
Sources
- Polar Sparsity: High Throughput Batched LLM Inferencing with Scalable Contextual Sparsity
- The $qs$ Inequality: Quantifying the Double Penalty of Mixture-of-Experts at Inference
- Characterization of Request and Token Energy Costs for LLM Inference Workloads on GPU Platforms
- From Tokens to Layers: Redefining Stall-Free Scheduling for MoE Serving with Layered Prefill
- Reconsidering the Energy Efficiency of Spiking Neural Networks Inference from Analytical Perspectives
- The ML.ENERGY Benchmark: Toward Automated Inference Energy Measurement and Optimization
- JaxMARL: Multi-Agent RL Environments and Algorithms in JAX
- Opportunistic Expert Activation: Batch-Aware Expert Routing for Faster Decode Without Retraining
- XShare: Collaborative in-Batch Expert Sharing for Faster MoE Inference
- Mixtral of Experts
- OLMoE: Open Mixture-of-Experts Language Models
- Qwen3 Technical Report
- MegaBlocks: Efficient Sparse Training with Mixture-of-Experts
- Scattered Mixture-of-Experts Implementation
- Lynx: Enabling Efficient MoE Inference through Dynamic Batch-Aware Expert Selection
- ReMoE: Boosting Expert Reuse through Router Fine-Tuning in Memory-Constrained MoE LLM Inference
- Sticky Routing: Training MoE Models for Memory-Efficient Inference
- Temporally Extended Mixture-of-Experts Models
- Not All Models Suit Expert Offloading: On Local Routing Consistency of Mixture-of-Expert Models
- Efficient recurrent architectures through activity sparsity and sparse back-propagation through time
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis