# DeltaLLM: Delta-Key Sparse Attention for Edge LLMs

arXiv: https://arxiv.org/abs/2507.19608 · HTML: https://mlsanity.com/paper/2507.19608/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 28 of its 33 references that are in the map.
- First posted on arXiv 25 Jul 2025.

## Builds on (most cited first, 28 of 28)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [HellaSwag: Can a Machine Really Finish Your Sentence?](https://mlsanity.com/paper/1905.07830/) (2019)
- [PIQA: Reasoning about Physical Commonsense in Natural Language](https://mlsanity.com/paper/1911.11641/) (2019)
- [Sparks of Artificial General Intelligence: Early experiments with GPT-4](https://mlsanity.com/paper/2303.12712/) (2023)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions](https://mlsanity.com/paper/1905.10044/) (2019)
- [Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answering](https://mlsanity.com/paper/1809.02789/) (2018)
- [Know What You Don't Know: Unanswerable Questions for SQuAD](https://mlsanity.com/paper/1806.03822/) (2018)
- [AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration](https://mlsanity.com/paper/2306.00978/) (2023)
- [SnapKV: LLM Knows What You are Looking for Before Generation](https://mlsanity.com/paper/2404.14469/) (2024)
- [Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads](https://mlsanity.com/paper/2401.10774/) (2024)
- [A Simple and Effective Pruning Approach for Large Language Models](https://mlsanity.com/paper/2306.11695/) (2023)
- [MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attention](https://mlsanity.com/paper/2407.02490/) (2024)
- [Unlocking Efficiency in Large Language Model Inference: A Comprehensive Survey of Speculative Decoding](https://mlsanity.com/paper/2401.07851/) (2024)
- [InfLLM: Training-Free Long-Context Extrapolation for LLMs with an Efficient Context Memory](https://mlsanity.com/paper/2402.04617/) (2024)
- [BitNet: Scaling 1-bit Transformers for Large Language Models](https://mlsanity.com/paper/2310.11453/) (2023)
- [FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference](https://mlsanity.com/paper/2502.20766/) (2025)
- [Training-Free Activation Sparsity in Large Language Models](https://mlsanity.com/paper/2408.14690/) (2024)
- [Skip-Convolutions for Efficient Video Processing](https://mlsanity.com/paper/2104.11487/) (2021)
- [Spartus: A 9.4 TOp/s FPGA-based LSTM Accelerator Exploiting Spatio-Temporal Sparsity](https://mlsanity.com/paper/2108.02297/) (2021)
- [EdgeDRNN: Recurrent Neural Network Accelerator for Edge Inference](https://mlsanity.com/paper/2012.13600/) (2020)
- [Delta Networks for Optimized Recurrent Network Computation](https://mlsanity.com/paper/1612.05571/) (2016)
- [DeltaCNN: End-to-End CNN Inference of Sparse Frame Differences in Videos](https://mlsanity.com/paper/2203.03996/) (2022)
- [MotionDeltaCNN: Sparse CNN Inference of Frame Differences in Moving Camera Videos](https://mlsanity.com/paper/2210.09887/) (2022)
- [Delta Keyword Transformer: Bringing Transformers to the Edge through Dynamically Pruned Multi-Head Self-Attention](https://mlsanity.com/paper/2204.03479/) (2022)
- [DeltaKWS: A 65nm 36nJ/Decision Bio-inspired Temporal-Sparsity-Aware Digital Keyword Spotting IC with 0.6V Near-Threshold SRAM](https://mlsanity.com/paper/2405.03905/) (2024)

## Built on by (most cited first, 5 of 5)

- [Attention Sink in Transformers: A Survey on Utilization, Interpretation, and Mitigation](https://mlsanity.com/paper/2604.10098/) (2026)
- [ResidualKV: Residual-Based KV Cache Compression for Efficient Long-Context Inference](https://mlsanity.com/paper/2602.08005/) (2026)
- [GLASS: Global-Local Aggregation for Inference-time Sparsification of LLMs](https://mlsanity.com/paper/2508.14302/) (2025)
- [On-Demand Multi-Task Sparsity for Efficient Large-Model Deployment on Edge Devices](https://mlsanity.com/paper/2511.19986/) (2025)
- [AEyeDE: An Attention-Based Attribution Framework for AI-Generated Text Detection](https://mlsanity.com/paper/2606.00016/) (2026)

## Abstract (arXiv)

On-device LLM inference is increasingly important for edge platforms but remains challenging due to high computational and memory demands and tight resource constraints. Modern applications often involve long prompts. Although compact models reduce linear-layer costs through weight compression, attention remains largely unoptimized and becomes the dominant bottleneck, especially during prefilling. Existing attention pruning methods primarily target highly parallel hardware (e.g., GPUs) and long context windows, limiting their suitability for edge scenarios. We present DeltaLLM, a training-free LLM framework based on algorithm-hardware-dataflow co-design to accelerate prefilling attention. DeltaLLM introduces a causal-attention-aware K-only delta formulation that enables Delta-Key reuse across query heads under Grouped-Query Attention (GQA). A context-aware hybrid strategy preserves exact attention near the diagonal while applying delta approximation to nonlocal regions. For score-value multiplication, we propose a vectorized structured sparse online softmax (VSSOS) that converts fine-grained spatial score sparsity into hardware-friendly column-level skipping. A specialized vector-processor architecture and unified sparse/dense attention dataflow exploit the resulting sparsity. Evaluated on BitNet-b1.58-2B-4T and LLaMA3.2-1B-Instruct, DeltaLLM achieves up to 60% Delta-Key sparsity and 63% structured attention-score sparsity with negligible average accuracy loss. On the KV260 FPGA board, it delivers a 1.62x speedup over TeLLMe-v2 while reducing DSP and BRAM usage. These results show that DeltaLLM supports efficient edge LLM deployment and integration with existing pipelines without fine-tuning.