# Accelerating Attention through Gradient-Based Learned Runtime Pruning

arXiv: https://arxiv.org/abs/2204.03227 · HTML: https://mlsanity.com/paper/2204.03227/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Builds on 52 of its 112 references that are in the map.
- First posted on arXiv 7 Apr 2022.

## Builds on (most cited first, 30 of 52)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Adam: A Method for Stochastic Optimization](https://mlsanity.com/paper/1412.6980/) (2014)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Graph Attention Networks](https://mlsanity.com/paper/1710.10903/) (2017)
- [SQuAD: 100,000+ Questions for Machine Comprehension of Text](https://mlsanity.com/paper/1606.05250/) (2016)
- [Longformer: The Long-Document Transformer](https://mlsanity.com/paper/2004.05150/) (2020)
- [Wide Residual Networks](https://mlsanity.com/paper/1605.07146/) (2016)
- [ALBERT: A Lite BERT for Self-supervised Learning of Language Representations](https://mlsanity.com/paper/1909.11942/) (2019)
- [Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention](https://mlsanity.com/paper/2006.16236/) (2020)
- [Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context](https://mlsanity.com/paper/1901.02860/) (2019)
- [Big Bird: Transformers for Longer Sequences](https://mlsanity.com/paper/2007.14062/) (2020)
- [Reformer: The Efficient Transformer](https://mlsanity.com/paper/2001.04451/) (2020)
- [Rethinking Attention with Performers](https://mlsanity.com/paper/2009.14794/) (2020)
- [Generating Long Sequences with Sparse Transformers](https://mlsanity.com/paper/1904.10509/) (2019)
- [Linformer: Self-Attention with Linear Complexity](https://mlsanity.com/paper/2006.04768/) (2020)
- [Are Sixteen Heads Really Better than One?](https://mlsanity.com/paper/1905.10650/) (2019)
- [In-Datacenter Performance Analysis of a Tensor Processing Unit](https://mlsanity.com/paper/1704.04760/) (2017)
- [Learning Sparse Neural Networks through L0 Regularization](https://mlsanity.com/paper/1712.01312/) (2017)
- [The State of Sparsity in Deep Neural Networks](https://mlsanity.com/paper/1902.09574/) (2019)
- [Efficient Content-Based Sparse Attention with Routing Transformers](https://mlsanity.com/paper/2003.05997/) (2020)
- [Towards AI-Complete Question Answering: A Set of Prerequisite Toy Tasks](https://mlsanity.com/paper/1502.05698/) (2015)
- [EIE: Efficient Inference Engine on Compressed Deep Neural Network](https://mlsanity.com/paper/1602.01528/) (2016)
- [SpAtten: Efficient Sparse Attention Architecture with Cascade Token and Head Pruning](https://mlsanity.com/paper/2012.09852/) (2020)
- [Structured Pruning of Large Language Models](https://mlsanity.com/paper/1910.04732/) (2019)
- [Chasing Sparsity in Vision Transformers: An End-to-End Exploration](https://mlsanity.com/paper/2106.04533/) (2021)

## Built on by (most cited first, 8 of 8)

- [Duplex: A Device for Large Language Models with Mixture of Experts, Grouped Query Attention, and Continuous Batching](https://mlsanity.com/paper/2409.01141/) (2024)
- [On Efficient Training of Large-Scale Deep Learning Models: A Literature Review](https://mlsanity.com/paper/2304.03589/) (2023)
- [Vision Transformers on the Edge: A Comprehensive Survey of Model Compression and Acceleration Strategies](https://mlsanity.com/paper/2503.02891/) (2025)
- [A Survey: Collaborative Hardware and Software Design in the Era of Large Language Models](https://mlsanity.com/paper/2410.07265/) (2024)
- [D'ej`a Vu: Efficient Video-Language Query Engine with Learning-based Inter-Frame Computation Reuse](https://mlsanity.com/paper/2506.14107/) (2025)
- [A Comprehensive Review and a Taxonomy of Edge Machine Learning: Requirements, Paradigms, and Techniques](https://mlsanity.com/paper/2302.08571/) (2023)
- [MCBP: A Memory-Compute Efficient LLM Inference Accelerator Leveraging Bit-Slice-enabled Sparsity and Repetitiveness](https://mlsanity.com/paper/2509.10372/) (2025)
- [BitStopper: An Efficient Transformer Attention Accelerator via Stage-fusion and Early Termination](https://mlsanity.com/paper/2512.06457/) (2025)

## Abstract (arXiv)

Self-attention is a key enabler of state-of-art accuracy for various transformer-based Natural Language Processing models. This attention mechanism calculates a correlation score for each word with respect to the other words in a sentence. Commonly, only a small subset of words highly correlates with the word under attention, which is only determined at runtime. As such, a significant amount of computation is inconsequential due to low attention scores and can potentially be pruned. The main challenge is finding the threshold for the scores below which subsequent computation will be inconsequential. Although such a threshold is discrete, this paper formulates its search through a soft differentiable regularizer integrated into the loss function of the training. This formulation piggy backs on the back-propagation training to analytically co-optimize the threshold and the weights simultaneously, striking a formally optimal balance between accuracy and computation pruning. To best utilize this mathematical innovation, we devise a bit-serial architecture, dubbed LeOPArd, for transformer language models with bit-level early termination microarchitectural mechanism. We evaluate our design across 43 back-end tasks for MemN2N, BERT, ALBERT, GPT-2, and Vision transformer models. Post-layout results show that, on average, LeOPArd yields 1.9x and 3.9x speedup and energy reduction, respectively, while keeping the average accuracy virtually intact (<0.2% degradation)