Characterizing High Bandwidth Flash for LLM Serving
cs.LG, cs.AR, cs.DC
Submitted: 2026-09-30
Updated: 2026-10-05
Code: https://github.com/deepseek-ai/LPLB
Sources
- IndexCache: Accelerating Sparse Attention via Cross-Layer Index Reuse
- DeepSeek-V3 Technical Report
- The Llama 3 Herd of Models
- HAVEN: High-Bandwidth Flash Augmented Vector Engine for Large-Scale Approximate Nearest-Neighbor Search Acceleration
- ThunderAgent: A Simple, Fast and Program-Aware Agentic Inference System
- Continuum: Efficient and Robust Multi-Turn LLM Agent Scheduling with KV Cache Time-to-Live
- HBF Sucks! A Full-Stack Characterization of High-Bandwidth Flash for KV-Centric LLM Serving
- LMCache: An Efficient KV Cache Layer for Enterprise-Scale LLM Inference
- Autellix: An Efficient Serving Engine for LLM Agents as General Programs
- Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism
- MemExplorer: Navigating the Heterogeneous Memory Design Space for Agentic Inference NPUs
- Qwen3 Technical Report
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks