LEAF: LUT-aware Efficient Attention on FPGAs
Abstract
We introduce Lookup-table (LUT)-aware Efficient Attention on FPGAs (LEAF). LEAF is the first nanosecond-scale LUT-aware transformer for FPGAs. Deploying transformers that are large enough to perform well, small enough to fit on an FPGA, and fast enough to meet the sub-microsecond timescales required by real-time systems such as those at CERN remains difficult, since the same attention mechanism that gives these models their high expressivity also makes them costly to implement in hardware. LEAF converts the query, key, and value projections and the final output feed-forward layer into FPGA-native layers with LUT-aware training, creating the first LUT-aware transformer and linformer layers built for ultra-low-latency inference. Attention resists LUT-aware training because it multiplies two activations where the method needs a constant weight to tabulate; LEAF tabulates the attention block's constant matrix-vector multiplications and leaves the rest in arithmetic form. On a standard jet tagging benchmark this improves accuracy by 1.8 percentage points while using 2.9 times fewer LUTs than the previous state-of-the-art sub-microsecond transformer at the same input size. Our most efficient LEAF-based linformer with 64 input particles reaches 80.4% classification accuracy at 28k LUTs and 35ns.
Related
- Hardware-aware quantum attention for fast radio burst identification
- ConCA: Concentration-Aware Channel Attention for Fine-Grained Visual Recognition
- Efficient angle-aware differentiable Mask3D modeling for inverse lithography
- Real-time Track Reconstruction with FPGAs: an example from MUonE
- C-Phase-Aware Compilation for Efficient Fault-Tolerant Quantum Execution
BibTeX
@misc{cox2026leaf,
title = {{LEAF: LUT-aware Efficient Attention on FPGAs}},
author = {Arianna Cox and Chang Sun and Lauri Laatu and Benedikt Maier and Aadeesh Sharma and Leo Rozanov and Wayne Luk and Maria Spiropulu and Alexander Tapper and Zhiqiang Que},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.27078},
eprint = {2610.27078},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.27078},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - LEAF: LUT-aware Efficient Attention on FPGAs
AU - Arianna Cox
AU - Chang Sun
AU - Lauri Laatu
AU - Benedikt Maier
AU - Aadeesh Sharma
AU - Leo Rozanov
AU - Wayne Luk
AU - Maria Spiropulu
AU - Alexander Tapper
AU - Zhiqiang Que
PY - 2026
DA - 2026/10/19
PB - arXiv
DO - 10.48550/arxiv.2610.27078
UR - https://arxiv.org/abs/2610.27078
AB - We introduce Lookup-table (LUT)-aware Efficient Attention on FPGAs (LEAF). LEAF is the first nanosecond-scale LUT-aware transformer for FPGAs. Deploying transformers that are large enough to perform well, small enough to fit on an FPGA, and fast enough to meet the sub-microsecond timescales required by real-time systems such as those at CERN remains difficult, since the same attention mechanism that gives these models their high expressivity also makes them costly to implement in hardware. LEAF converts the query, key, and value projections and the final output feed-forward layer into FPGA-native layers with LUT-aware training, creating the first LUT-aware transformer and linformer layers built for ultra-low-latency inference. Attention resists LUT-aware training because it multiplies two activations where the method needs a constant weight to tabulate; LEAF tabulates the attention block's constant matrix-vector multiplications and leaves the rest in arithmetic form. On a standard jet tagging benchmark this improves accuracy by 1.8 percentage points while using 2.9 times fewer LUTs than the previous state-of-the-art sub-microsecond transformer at the same input size. Our most efficient LEAF-based linformer with 64 input particles reaches 80.4% classification accuracy at 28k LUTs and 35ns.
ER -