Saved in:
Bibliographic Details
Main Authors: Shen, Jeff, Smith, Lindsay M.
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.07282
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914055529168896
author Shen, Jeff
Smith, Lindsay M.
author_facet Shen, Jeff
Smith, Lindsay M.
contents We present cryptogram solving as an ideal testbed for studying neural network reasoning and generalization; models must decrypt text encoded with substitution ciphers, choosing from 26! possible mappings without explicit access to the cipher. We develop ALICE (an Architecture for Learning Interpretable Cryptogram dEcipherment), a simple encoder-only Transformer that sets a new state-of-the-art for both accuracy and speed on this decryption problem. Surprisingly, ALICE generalizes to unseen ciphers after training on only ${\sim}1500$ unique ciphers, a minute fraction ($3.7 \times 10^{-24}$) of the possible cipher space. To enhance interpretability, we introduce a novel bijective decoding head that explicitly models permutations via the Gumbel-Sinkhorn method, enabling direct extraction of learned cipher mappings. Through early exit and probing experiments, we reveal how ALICE progressively refines its predictions in a way that appears to mirror common human strategies -- early layers place greater emphasis on letter frequencies, while later layers form word-level structures. Our architectural innovations and analysis methods are applicable beyond cryptograms and offer new insights into neural network generalization and interpretability.
format Preprint
id arxiv_https___arxiv_org_abs_2509_07282
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ALICE: An Interpretable Neural Architecture for Generalization in Substitution Ciphers
Shen, Jeff
Smith, Lindsay M.
Machine Learning
Artificial Intelligence
Computation and Language
Cryptography and Security
We present cryptogram solving as an ideal testbed for studying neural network reasoning and generalization; models must decrypt text encoded with substitution ciphers, choosing from 26! possible mappings without explicit access to the cipher. We develop ALICE (an Architecture for Learning Interpretable Cryptogram dEcipherment), a simple encoder-only Transformer that sets a new state-of-the-art for both accuracy and speed on this decryption problem. Surprisingly, ALICE generalizes to unseen ciphers after training on only ${\sim}1500$ unique ciphers, a minute fraction ($3.7 \times 10^{-24}$) of the possible cipher space. To enhance interpretability, we introduce a novel bijective decoding head that explicitly models permutations via the Gumbel-Sinkhorn method, enabling direct extraction of learned cipher mappings. Through early exit and probing experiments, we reveal how ALICE progressively refines its predictions in a way that appears to mirror common human strategies -- early layers place greater emphasis on letter frequencies, while later layers form word-level structures. Our architectural innovations and analysis methods are applicable beyond cryptograms and offer new insights into neural network generalization and interpretability.
title ALICE: An Interpretable Neural Architecture for Generalization in Substitution Ciphers
topic Machine Learning
Artificial Intelligence
Computation and Language
Cryptography and Security
url https://arxiv.org/abs/2509.07282