# SwitchHead: Accelerating Transformers with Mixture-of-Experts Attention

arXiv: https://arxiv.org/abs/2312.07987 · HTML: https://mlsanity.com/paper/2312.07987/

## Key facts

- Cited by 34 papers in the map (data through 14 Oct 2026).
- Placed in Expert routing strategies, one of 12 directions in Mixture-of-experts.
- Builds on 32 of its 46 references that are in the map.
- First posted on arXiv 13 Dec 2023.

## Placed in

- Mixture-of-experts / [Expert routing strategies](https://mlsanity.com/moe/expert-routing-strategies/)

## Builds on (most cited first, 30 of 32)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [Neural Machine Translation by Jointly Learning to Align and Translate](https://mlsanity.com/paper/1409.0473/) (2014)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer](https://mlsanity.com/paper/1701.06538/) (2017)
- [Pointer Sentinel Mixture Models](https://mlsanity.com/paper/1609.07843/) (2016)
- [Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity](https://mlsanity.com/paper/2101.03961/) (2021)
- [Neural Machine Translation of Rare Words with Subword Units](https://mlsanity.com/paper/1508.07909/) (2015)
- [Sparks of Artificial General Intelligence: Early experiments with GPT-4](https://mlsanity.com/paper/2303.12712/) (2023)
- [Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention](https://mlsanity.com/paper/2006.16236/) (2020)
- [SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing](https://mlsanity.com/paper/1808.06226/) (2018)
- [Transformer-XL: Attentive Language Models Beyond a Fixed-Length Context](https://mlsanity.com/paper/1901.02860/) (2019)
- [GShard: Scaling Giant Models with Conditional Computation and Automatic Sharding](https://mlsanity.com/paper/2006.16668/) (2020)
- [The LAMBADA dataset: Word prediction requiring a broad discourse context](https://mlsanity.com/paper/1606.06031/) (2016)
- [In-context Learning and Induction Heads](https://mlsanity.com/paper/2209.11895/) (2022)
- [Fast Transformer Decoding: One Write-Head is All You Need](https://mlsanity.com/paper/1911.02150/) (2019)
- [BLiMP: The Benchmark of Linguistic Minimal Pairs for English](https://mlsanity.com/paper/1912.00582/) (2019)
- [Linear Transformers Are Secretly Fast Weight Programmers](https://mlsanity.com/paper/2102.11174/) (2021)
- [Random Feature Attention](https://mlsanity.com/paper/2103.02143/) (2021)
- [BASE Layers: Simplifying Training of Large, Sparse Models](https://mlsanity.com/paper/2103.16716/) (2021)
- [Unified Scaling Laws for Routed Language Models](https://mlsanity.com/paper/2202.01169/) (2022)
- [On the Representation Collapse of Sparse Mixture of Experts](https://mlsanity.com/paper/2204.09179/) (2022)
- [ListOps: A Diagnostic Dataset for Latent Tree Learning](https://mlsanity.com/paper/1804.06028/) (2018)
- [The Goldilocks Principle: Reading Children's Books with Explicit Memory Representations](https://mlsanity.com/paper/1511.02301/) (2015)
- [The Neural Data Router: Adaptive Control Flow in Transformers Improves Systematic Generalization](https://mlsanity.com/paper/2110.07732/) (2021)
- [Mixture of Attention Heads: Selecting Attention Heads Per Token](https://mlsanity.com/paper/2210.05144/) (2022)

## Built on by (most cited first, 30 of 34)

- [Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free](https://mlsanity.com/paper/2505.06708/) (2025)
- [Towards Efficient Generative Large Language Model Serving: A Survey from Algorithms to Systems](https://mlsanity.com/paper/2312.15234/) (2023)
- [A Comprehensive Survey of Mixture-of-Experts: Algorithms, Theory, and Applications](https://mlsanity.com/paper/2503.07137/) (2025)
- [Model Compression and Efficient Inference for Large Language Models: A Survey](https://mlsanity.com/paper/2402.09748/) (2024)
- [MoEUT: Mixture-of-Experts Universal Transformers](https://mlsanity.com/paper/2405.16039/) (2024)
- [ReMoE: Fully Differentiable Mixture-of-Experts with ReLU Routing](https://mlsanity.com/paper/2412.14711/) (2024)
- [A Survey on Inference Optimization Techniques for Mixture of Experts Models](https://mlsanity.com/paper/2412.14219/) (2024)
- [MoH: Multi-Head Attention as Mixture-of-Head Attention](https://mlsanity.com/paper/2410.11842/) (2024)
- [BAM! Just Like That: Simple and Efficient Parameter Upcycling for Mixture of Experts](https://mlsanity.com/paper/2408.08274/) (2024)
- [Scattered Mixture-of-Experts Implementation](https://mlsanity.com/paper/2403.08245/) (2024)
- [Mixture of Sparse Attention: Content-Based Learnable Sparse Attention via Expert-Choice Routing](https://mlsanity.com/paper/2505.00315/) (2025)
- [Sparse Layers are Critical to Scaling Looped Language Models](https://mlsanity.com/paper/2605.09165/) (2026)
- [Learning to Skip the Middle Layers of Transformers](https://mlsanity.com/paper/2506.21103/) (2025)
- [UMoE: Unifying Attention and FFN with Shared Experts](https://mlsanity.com/paper/2505.07260/) (2025)
- [Sigmoid Self-Attention has Lower Sample Complexity than Softmax Self-Attention: A Mixture-of-Experts Perspective](https://mlsanity.com/paper/2502.00281/) (2025)
- [Routing Mamba: Scaling State Space Models with Mixture-of-Experts Projection](https://mlsanity.com/paper/2506.18145/) (2025)
- [Sparse Growing Transformer: Training-Time Sparse Depth Allocation via Progressive Attention Looping](https://mlsanity.com/paper/2603.23998/) (2026)
- [MoRE: Mixture of Reused Experts](https://mlsanity.com/paper/2609.18176/) (2026)
- [BLR-MoE: Boosted Language-Routing Mixture of Experts for Domain-Robust Multilingual E2E ASR](https://mlsanity.com/paper/2501.12602/) (2025)
- [Reasoning-Intensive Regression](https://mlsanity.com/paper/2508.21762/) (2025)
- [DoRAN: Stabilizing Weight-Decomposed Low-Rank Adaptation via Noise Injection and Auxiliary Networks](https://mlsanity.com/paper/2510.04331/) (2025)
- [Grouped Query Experts: Mixture-of-Experts on GQA Self-Attention](https://mlsanity.com/paper/2606.20945/) (2026)
- [Mixture of Group Experts for Learning Invariant Representations](https://mlsanity.com/paper/2504.09265/) (2025)
- [LiMoDE: Rethinking Lifelong Robot Manipulation from a Mixture-of-Dynamic-Experts Perspective](https://mlsanity.com/paper/2606.26183/) (2026)
- [RouterKT: Mixture-of-Experts for Knowledge Tracing](https://mlsanity.com/paper/2504.08989/) (2025)
- [Efficient Inter-Task Attention for Multitask Transformer Models](https://mlsanity.com/paper/2508.04422/) (2025)
- [Mixture-of-Transformers Learn Faster: A Theoretical Study on Classification Problems](https://mlsanity.com/paper/2510.27004/) (2025)
- [HIMOSA: Efficient Remote Sensing Image Super-Resolution with Hierarchical Mixture of Sparse Attention](https://mlsanity.com/paper/2512.00275/) (2025)
- [Non-Interfering Weight Fields: Treating Model Parameters as a Continuously Extensible Function](https://mlsanity.com/paper/2602.18628/) (2026)
- [Swimba: Switch Mamba Model Scales State Space Models](https://mlsanity.com/paper/2603.06938/) (2026)

## Abstract (arXiv)

Despite many recent works on Mixture of Experts (MoEs) for resource-efficient Transformer language models, existing methods mostly focus on MoEs for feedforward layers. Previous attempts at extending MoE to the self-attention layer fail to match the performance of the parameter-matched baseline. Our novel SwitchHead is an effective MoE method for the attention layer that successfully reduces both the compute and memory requirements, achieving wall-clock speedup, while matching the language modeling performance of the baseline Transformer. Our novel MoE mechanism allows SwitchHead to compute up to 8 times fewer attention matrices than the standard Transformer. SwitchHead can also be combined with MoE feedforward layers, resulting in fully-MoE "SwitchAll" Transformers. For our 262M parameter model trained on C4, SwitchHead matches the perplexity of standard models with only 44% compute and 27% memory usage. Zero-shot experiments on downstream tasks confirm the performance of SwitchHead, e.g., achieving more than 3.5% absolute improvements on BliMP compared to the baseline with an equal compute resource.