DeltaLLM: Delta-Key Sparse Attention for Edge LLMs

arXiv:2507.19608 · cs.AI, eess.SP · Submitted 2025-07-25 · Read on arXiv

cs.AI, eess.SP

Submitted: 2025-07-25

Updated: 2026-10-09

Comments: This paper has been accepted by IEEE Transactions on Circuits and Systems for Artificial Intelligence (TCASAI)

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

Terminology

• Accuracy — 4× in this paper · explained in 1 episode
• Sparsity — 4× in this paper · explained in 4 episodes
• Edge Devices — 2× in this paper · explained in 2 episodes
• Attention Mechanism — 1× in this paper · explained in 3 episodes
• Edge Deployment — 1× in this paper · explained in 2 episodes
• F1 Score — 1× in this paper · explained in 3 episodes
• Fine-Tuning — 1× in this paper · explained in 5 episodes
• Large Language Models (LLMs) — 1× in this paper · explained in 8 episodes
• Pruning — 1× in this paper · explained in 4 episodes
• Sparse Attention — 1× in this paper · explained in 5 episodes
• Training-Free Framework — 1× in this paper · explained in 2 episodes

Sources

• LLaMA: Open and Efficient Foundation Language Models
• Training-Free Activation Sparsity in Large Language Models
• Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads
• Unlocking Efficiency in Large Language Model Inference: A Comprehensive Survey of Speculative Decoding
• A Simple and Effective Pruning Approach for Large Language Models
• How Sparse Attention Approximates Exact Attention? Your Attention is Naturally $n^C$-Sparse
• Efficient Streaming Language Models with Attention Sinks
• MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attention
• InfLLM: Training-Free Long-Context Extrapolation for LLMs with an Efficient Context Memory
• FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference
• SpargeAttention: Accurate and Training-free Sparse Attention Accelerating Any Model Inference
• Delta Keyword Transformer: Bringing Transformers to the Edge through Dynamically Pruned Multi-Head Self-Attention
• The Llama 3 Herd of Models
• BitNet: Scaling 1-bit Transformers for Large Language Models
• Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
• BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions
• HellaSwag: Can a Machine Really Finish Your Sentence?
• PIQA: Reasoning about Physical Commonsense in Natural Language

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• Agents' Last Exam
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems