Long-Context Modeling with Dynamic Hierarchical Sparse Attention for On-Device LLMs

TL;DR

This paper introduces Dynamic Hierarchical Sparse Attention (DHSA), a novel, data-driven method for long-context modeling in resource-limited on-device LLMs that improves efficiency and accuracy without retraining.

Contribution

DHSA adaptively segments sequences and computes importance scores dynamically, outperforming static and heuristic methods in accuracy and efficiency for long-context LLMs.

Findings

DHSA reduces latency by 20-60% and memory by 35%.

DHSA achieves 6-18% higher accuracy than baseline methods.

DHSA matches dense attention accuracy with lower computational cost.

Abstract

The quadratic cost of attention hinders the scalability of long-context LLMs, especially in resource-constrained settings. Existing static sparse methods such as sliding windows or global tokens utilizes the sparsity of attention to reduce the cost of attention, but poorly adapts to the content-dependent variations in attention due to their staticity. While previous work has proposed several dynamic approaches to improve flexibility, they still depend on predefined templates or heuristic mechanisms. Such strategies reduce generality and prune tokens that remain contextually important, limiting their accuracy across diverse tasks. To tackle these bottlenecks of existing methods for long-context modeling, we introduce Dynamic Hierarchical Sparse Attention (DHSA), a data-driven framework that dynamically predicts attention sparsity online without retraining. Our proposed DHSA adaptively segments sequences into variable-length chunks, then computes chunk representations by aggregating the token embeddings within each chunk. To avoid the bias introduced by varying chunk lengths, we apply length-normalized aggregation that scales the averaged embeddings by the square root of the chunk size. Finally, DHSA upsamples the chunk-level similarity scores to token level similarities to calculate importance scores that determine which token-level interactions should be preserved. Our experiments on Gemma2 with Needle-in-a-Haystack Test and LongBench show that DHSA matches dense attention in accuracy, while reducing prefill latency by 20-60% and peak memory usage by 35%. Compared to other representative baselines such as block sparse attention, DHSA achieves consistently higher accuracy (6-18% relative gains) with comparable or lower cost, offering an efficient and adaptable solution for long-context on-device LLMs.