mlsanityData to 7 Oct 2026

From Arithmetic to Logic: The Resilience of Logic and Lookup-Based Neural Networks Under Parameter Bit-Flips

arXiv 2603.22770 ↗First posted 24 Mar 2026cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

The deployment of deep neural networks (DNNs) in safety-critical edge environments necessitates robustness against hardware-induced bit-flip errors. While empirical studies indicate that reducing numerical precision can improve fault tolerance, the theoretical basis of this phenomenon remains underexplored. In this work, we study resilience as a structural property of neural architectures rather than solely as a property of a dataset-specific trained solution. By deriving the expected squared error (MSE) under independent parameter bit flips across multiple numerical formats and layer primitives, we show that lower precision, higher sparsity, bounded activations, and shallow depth are consistently favored under this corruption model. We then argue that logic and lookup-based neural networks realize the joint limit of these design trends. Through ablation studies on the MLPerf Tiny benchmark suite, we show that the observed empirical trends are consistent with the theoretical predictions, and that LUT-based models remain highly stable in corruption regimes where standard floating-point models fail sharply. Furthermore, we identify a novel even-layer recovery effect unique to logic-based architectures and analyze the structural conditions under which it emerges. Overall, our results suggest that shifting from continuous arithmetic weights to discrete Boolean lookups can provide a favorable accuracy-resilience trade-off for hardware fault tolerance.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 31 of its 55 references that are in the map. First posted on arXiv 24 Mar 2026.

Builds on

31 of 55 references.

References · 31

  1. 2017Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithmscited by 4,283
  2. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,031
  3. 2017Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inferencecited by 1,548
  4. 2016Quantized Neural Networks: Training Neural Networks with Low Precision Weights and Activationscited by 413
  5. 2019Understanding Straight-Through Estimator in Training Activation Quantized Neural Netscited by 225
  6. 2021MLPerf Tiny Benchmarkcited by 118
  7. 2019Bit-Flip Attack: Crushing Neural Network with Progressive Bit Searchcited by 106
  8. 2022Deep Differentiable Logic Gate Networkscited by 67
  9. 2020LogicNets: Co-Designed Neural Networks and Circuits for Extreme-Throughput Applicationscited by 55
  10. 2024Convolutional Differentiable Logic Gate Networkscited by 37
  11. 2023PolyLUT: Learning Piecewise Polynomials for Ultra-Low Latency FPGA LUT-based Inferencecited by 30
  12. 2020EMPIR: Ensembles of Mixed Precision Deep Networks for Increased Robustness against Adversarial Attackscited by 28
19 more
  1. 2024Differentiable Weightless Neural Networkscited by 27
  2. 2020Least squares binary quantization of neural networkscited by 21
  3. 2025TreeLUT: An Efficient Alternative to Deep Neural Networks for Inference Acceleration Using Gradient Boosted Decision Treescited by 15
  4. 2025NeuraLUT-Assemble: Hardware-aware Assembling of Sub-Neural Networks for Efficient LUT Inferencecited by 14
  5. 2023Resilient Binary Neural Networkcited by 13
  6. 2025KANEL'E: Kolmogorov-Arnold Networks for Efficient LUT-based Evaluationcited by 13
  7. 2025Light Differentiable Logic Gate Networkscited by 9
  8. 2023ULEEN: A Novel Architecture for Ultra Low-Energy Edge Neural Networkscited by 8
  9. 2025On Jailbreaking Quantized Language Models Through Fault Injection Attackscited by 8
  10. 2024ReducedLUT: Table Decomposition with "Don't Care" Conditionscited by 7
  11. 2025Loss Landscape Analysis for Reliable Quantized ML Models for Scientific Sensingcited by 3
  12. 2025LL-ViT: Edge Deployable Vision Transformers with Look Up Table Neuronscited by 3
  13. 2021Hardware-friendly Deep Learning by Network Quantization and Binarizationcited by 2
  14. 2023Compiled Models, Built-In Exploits: Uncovering Pervasive Bit-Flip Attack Surfaces in DNN Executablescited by 2
  15. 2025Verification of Bit-Flip Attacks against Quantized Neural Networkscited by 2
  16. 2025SparseLUT: Sparse Connectivity Optimization for Lookup Table-based Deep Neural Networkscited by 2
  17. 2026WARP Logic Neural Networkscited by 2
  18. 2024ZOBNN: Zero-Overhead Dependable Design of Binary Neural Networks with Deliberately Quantized Parameterscited by 1
  19. 2025Differentiable Weightless Controllers: Learning Logic Circuits for Continuous Controlcited by 1

Built on by

Not cited yet.