RouteScan: A Non-Intrusive Approach to Auditing MoE LLMs Safety via Expert Routing Telemetry

Documenti analoghi

• MoE-GPS: Guidlines for Prediction Strategy for Dynamic Expert Duplication in MoE Load Balancing
  di: Ma, Haiyue, et al.
  Pubblicazione: (2025)
• SliceMoE: Bit-Sliced Expert Caching under Miss-Rate Constraints for Efficient MoE Inference
  di: Choi, Yuseon, et al.
  Pubblicazione: (2025)
• A3D-MoE: Acceleration of Large Language Models with Mixture of Experts via 3D Heterogeneous Integration
  di: Huang, Wei-Hsing, et al.
  Pubblicazione: (2025)
• Expert Streaming: Accelerating Low-Batch MoE Inference via Multi-chiplet Architecture and Dynamic Expert Trajectory Scheduling
  di: Ma, Songchen, et al.
  Pubblicazione: (2026)
• Accelerating Detailed Routing Convergence through Offline Reinforcement Learning
  di: Khan, Afsara, et al.
  Pubblicazione: (2025)