A permutation-equivariant transformer trained on self-supervised oracle trajectories from scrambled expressions achieves near-perfect simplification rates for dilogarithms and 100% success on 5-point gluon scattering amplitudes with over 200 terms.
The similarity threshold starts at 0.6 and relaxes across passes (through 0.5 and 0.4), progressively allowing less similar terms to be grouped
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
hep-th 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Learning to Unscramble: Simplifying Symbolic Expressions via Self-Supervised Oracle Trajectories
A permutation-equivariant transformer trained on self-supervised oracle trajectories from scrambled expressions achieves near-perfect simplification rates for dilogarithms and 100% success on 5-point gluon scattering amplitudes with over 200 terms.