ATFlash: Per-RoPE-Wavelength Attention Windows for Compute/Memory-Efficient LLM Inference

arXiv:2608.02947 · cs.LG, cs.CL · Submitted 2026-08-03 · Read on arXiv

Shun-ichiro Hayashi, Daichi Mukunoki, Tetsuya Hoshino, Takahiro Katagiri

cs.LG, cs.CL

Submitted: 2026-08-03

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Decode — 33× in this paper · explained in 2 episodes
• Pruning — 14× in this paper · explained in 4 episodes
• Benchmark — 11× in this paper · explained in 2 episodes
• Accuracy — 10× in this paper · explained in 1 episode
• Sparse Attention — 5× in this paper · explained in 5 episodes
• Verification — 5× in this paper · explained in 2 episodes
• In International Conference on Learning Representations — 4× in this paper
• Fidelity — 3× in this paper · explained in 2 episodes
• KV cache — 3× in this paper · explained in 3 episodes
• Sparsity — 3× in this paper · explained in 4 episodes
• Transformer — 3× in this paper · explained in 2 episodes
• Core — 2× in this paper · explained in 2 episodes
• Embeddings — 2× in this paper · explained in 2 episodes
• In International Conference on Machine Learning — 2× in this paper
• Adaptation — 1× in this paper · explained in 1 episode
• Attention Mechanism — 1× in this paper · explained in 3 episodes
• Edge AI — 1× in this paper · explained in 2 episodes
• Generalization — 1× in this paper · explained in 13 episodes
• Hybrid Architecture — 1× in this paper · explained in 2 episodes
• Large Language Models (LLMs) — 1× in this paper · explained in 8 episodes
• Model-Agnostic — 1× in this paper · explained in 2 episodes
• Quantization — 1× in this paper · explained in 14 episodes
• Robustness — 1× in this paper · explained in 5 episodes
• Trace — 1× in this paper · explained in 2 episodes
• Transformers — 1× in this paper · explained in 2 episodes

Sources

• Longformer: The Long-Document Transformer
• DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
• Flex Attention: A Programming Model for Generating Optimized Attention Kernels
• TriAttention: Efficient Long Reasoning with Trigonometric KV Compression
• Online normalizer calculation for softmax
• MEPIC: Memory Efficient Position Independent Caching for LLM Serving
• Scaling Attention via Feature Sparsity
• FlashAttention-4: Algorithm and Kernel Pipelining Co-Design for Asymmetric Hardware Scaling

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