mlsanityData to 7 Oct 2026

Algebraic Invariants of Lightning Self-Attention

arXiv 2604.15632 ↗First posted 17 Apr 2026math.AG stat.MLcited by 1

Not on the map. Why?

Abstract (arXiv)

We study the polynomial coefficients of lightning self-attention as coordinates of an algebraic variety. In the single-token case, we characterize the coefficient variety as a rank-constrained Chow-type variety and derive algebraic equations from this description. For multiple tokens, linear relations reduce the nonlinear geometry to coefficients involving interactions between distinct tokens. We characterize the resulting variety by a common linear factor together with a low-rank condition on the remaining factors. This structure yields explicit families of determinantal, Veronese-type, and Sylvester resultant-based invariants. In the rank-one case, natural pencil and flattening equations define the variety set-theoretically. We complement these results with computations in small dimensions, where the theoretical constructions recover the generators of the defining ideals. Overall, the paper gives a structural description of the algebraic constraints imposed on polynomial coefficient arrays by lightning self-attention and provides explicit certificates of non-realizability.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 31 of its 57 references that are in the map. First posted on arXiv 17 Apr 2026.

Builds on

31 of 57 references.

References · 31

  1. 2017Attention Is All You Needcited by 59,932
  2. 2017Reluplex: An Efficient SMT Solver for Verifying Deep Neural Networkscited by 739
  3. 2017Provable defenses against adversarial examples via the convex outer adversarial polytopecited by 564
  4. 2019Are Transformers universal approximators of sequence-to-sequence functions?cited by 404
  5. 2017Formal Verification of Piece-Wise Linear Feed-Forward Neural Networkscited by 241
  6. 2021Beta-CROWN: Efficient Bound Propagation with Per-neuron Split Constraints for Complete and Incomplete Neural Network Robustness Verificationcited by 224
  7. 2020The Lipschitz Constant of Self-Attentioncited by 171
  8. 2023A mathematical perspective on Transformerscited by 160
  9. 2019Branch and Bound for Piecewise Linear Neural Network Verificationcited by 138
  10. 2018A Dual Approach to Scalable Verification of Deep Networkscited by 125
  11. 2018Tropical Geometry of Deep Neural Networkscited by 92
  12. 2019On the Expressive Power of Deep Polynomial Neural Networkscited by 65
19 more
  1. 2021Introduction to Neural Network Verificationcited by 63
  2. 2022Your Transformer May Not be as Powerful as You Expectcited by 55
  3. 2020A Mathematical Theory of Attentioncited by 43
  4. 2023What can a Single Attention Layer Learn? A Study Through the Random Features Lenscited by 39
  5. 2019Pure and Spurious Critical Points: a Geometric Study of Linear Networkscited by 33
  6. 2021Geometry of Linear Convolutional Networkscited by 27
  7. 2024The Real Tropical Geometry of Neural Networkscited by 23
  8. 2024Geometry of Polynomial Neural Networkscited by 21
  9. 2025Algebra Unveils Deep Learning -- An Invitation to Neuroalgebraic Geometrycited by 21
  10. 2023Function Space and Critical Points of Linear Convolutional Networkscited by 19
  11. 2024Geometry of Lightning Self-Attention: Identifiability and Dimensioncited by 16
  12. 2024Activation degree thresholds and expressiveness of polynomial neural networkscited by 14
  13. 2025Identifiability of Deep Polynomial Neural Networkscited by 12
  14. 2024Algebraic Complexity and Neurovariety of Linear Convolutional Networkscited by 10
  15. 2015Dimension of Marginals of Kronecker Product Modelscited by 4
  16. 2017Mixtures and products in two graphical modelscited by 4
  17. 2024Attention is a smoothed cubic splinecited by 4
  18. 2025Constraining the outputs of ReLU neural networkscited by 3
  19. 2026Robustness Verification of Polynomial Neural Networkscited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.