TileWeave: Heterogeneous Tiling for Sparse Prefill on Mobile SoCs
cs.LG
Submitted: 2026-10-30
Updated: 2026-10-30
Code: https://github.com/ggml-org/llama
Terminology
Sources
- Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone
- LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding
- FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness
- FlashPrefill: Instantaneous Pattern Discovery and Thresholding for Ultra-Fast Long-Context Prefilling
- SeerAttention: Learning Intrinsic Sparse Attention in Your LLMs
- Scaling LLM Test-Time Compute with Mobile NPU on Smartphones
- RULER: What's the Real Context Size of Your Long-Context Language Models?
- MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attention
- FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference
- Twilight: Adaptive Attention Sparsity with Hierarchical Top-$p$ Pruning
- MoBA: Mixture of Block Attention for Long-Context LLMs
- Online normalizer calculation for softmax
- Double-P: Hierarchical Top-P Sparse Attention for Long-Context LLMs
- Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inference
- Fast On-device LLM Inference with NPUs
- XAttention: Block Sparse Attention with Antidiagonal Scoring
- PowerInfer-2: Fast Large Language Model Inference on a Smartphone
- Qwen3 Technical Report
- ShadowNPU: System and Algorithm Co-design for NPU-Centric On-Device LLM Inference
- SpargeAttention: Accurate and Training-free Sparse Attention Accelerating Any Model Inference
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