# COBRA: Algorithm-Architecture Co-optimized Binary Transformer Accelerator for Edge Inference

arXiv: https://arxiv.org/abs/2504.16269 · HTML: https://mlsanity.com/paper/2504.16269/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- Builds on 19 of its 26 references that are in the map.
- First posted on arXiv 22 Apr 2025.

## Builds on (most cited first, 19 of 19)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [GLUE: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding](https://mlsanity.com/paper/1804.07461/) (2018)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [ProgPrompt: Generating Situated Robot Task Plans using Large Language Models](https://mlsanity.com/paper/2209.11302/) (2022)
- [Q8BERT: Quantized 8Bit BERT](https://mlsanity.com/paper/1910.06188/) (2019)
- [BinaryBERT: Pushing the Limit of BERT Quantization](https://mlsanity.com/paper/2012.15701/) (2020)
- [A Fast Post-Training Pruning Framework for Transformers](https://mlsanity.com/paper/2204.09656/) (2022)
- [TernaryBERT: Distillation-aware Ultra-low Bit BERT](https://mlsanity.com/paper/2009.12812/) (2020)
- [BiBERT: Accurate Fully Binarized BERT](https://mlsanity.com/paper/2203.06390/) (2022)
- [BiT: Robustly Binarized Multi-distilled Transformer](https://mlsanity.com/paper/2205.13016/) (2022)
- [FracBNN: Accurate and FPGA-Efficient Binary Neural Networks with Fractional Activations](https://mlsanity.com/paper/2012.12206/) (2020)
- [VAQF: Fully Automatic Software-Hardware Co-Design Framework for Low-Bit Vision Transformer](https://mlsanity.com/paper/2201.06618/) (2022)
- [Hardware Acceleration of Fully Quantized BERT for Efficient Natural Language Processing](https://mlsanity.com/paper/2103.02800/) (2021)
- [A Two-Stage Efficient 3-D CNN Framework for EEG Based Emotion Recognition](https://mlsanity.com/paper/2208.00883/) (2022)
- [BETA: Binarized Energy-Efficient Transformer Accelerator at the Edge](https://mlsanity.com/paper/2401.11851/) (2024)
- [XNORBIN: A 95 TOp/s/W Hardware Accelerator for Binary Convolutional Neural Networks](https://mlsanity.com/paper/1803.05849/) (2018)
- [Support for Stock Trend Prediction Using Transformers and Sentiment Analysis](https://mlsanity.com/paper/2305.14368/) (2023)
- [Co-Designing Binarized Transformer and Hardware Accelerator for Efficient End-to-End Edge Deployment](https://mlsanity.com/paper/2407.12070/) (2024)

## Built on by (most cited first, 4 of 4)

- [TeLLMe v2: An Efficient End-to-End Ternary LLM Prefill and Decode Accelerator with Table-Lookup Matmul on Edge FPGAs](https://mlsanity.com/paper/2510.15926/) (2025)
- [PD-Swap: Prefill-Decode Logic Swapping for End-to-End LLM Inference on Edge FPGAs via Dynamic Partial Reconfiguration](https://mlsanity.com/paper/2512.11550/) (2025)
- [FASQ: Flexible Accelerated Subspace Quantization for Calibration-Free LLM Compression](https://mlsanity.com/paper/2605.04084/) (2026)
- [Recent Developments in Transformer Inference Deployment on FPGA Platforms: A Survey](https://mlsanity.com/paper/2609.01212/) (2026)

## Abstract (arXiv)

Transformer-based models have demonstrated superior performance in various fields, including natural language processing and computer vision. However, their enormous model size and high demands in computation, memory, and communication limit their deployment to edge platforms for local, secure inference. Binary transformers offer a compact, low-complexity solution for edge deployment with reduced bandwidth needs and acceptable accuracy. However, existing binary transformers perform inefficiently on current hardware due to the lack of binary specific optimizations. To address this, we introduce COBRA, an algorithm-architecture co-optimized binary Transformer accelerator for edge computing. COBRA features a real 1-bit binary multiplication unit, enabling matrix operations with -1, 0, and +1 values, surpassing ternary methods. With further hardware-friendly optimizations in the attention block, COBRA achieves up to 3,894.7 GOPS throughput and 448.7 GOPS/Watt energy efficiency on edge FPGAs, delivering a 311x energy efficiency improvement over GPUs and a 3.5x throughput improvement over the state-of-the-art binary accelerator, with only negligible inference accuracy degradation.