Newest papers

Hierarchical Token Merging with Recursive Autoencoders for Unsupervised Grammar Induction

Murat Guney Kemal, Robert Lieck · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26695 · cited by 0 works (OpenAlex) · open-access PDF

Spanish

El aprendizaje de tales jerarquías directamente de los datos, sin supervisión estructural explícita, sigue siendo un desafío central en el aprendizaje de la representación. Investigamos si una estructura jerárquica significativa puede surgir de la compresión sola. Mientras que los autoencoders aprenden las representaciones comprimidas a través de la reconstrucción, este objetivo no favorece necesariamente la estructura de la composición significativa. Estudiamos este desafío a través de una inducción gramatical no supervisada, donde la jerarquía subyacente puede ser evaluada explícitamente. Introducimos un autoencodificador jerárquico de fusión de tokens que combina de forma recursiva los tokens adyacentes en representaciones cada vez más abstractas, comprimiendo secuencias enteras a través de una jerarquía de información variable. El modelo

Machine translation (Meta M2M100); the English original follows.

Abstract

Hierarchical abstraction is a fundamental principle across many data modalities, enabling complex sequences to be represented compositionally at multiple levels. Learning such hierarchies directly from data, without explicit structural supervision, remains a central challenge in representation learning. We investigate whether meaningful hierarchical structure can emerge from compression alone. While autoencoders learn compressed representations through reconstruction, this objective does not necessarily favour meaningful compositional structure. We study this challenge through unsupervised grammar induction, where the underlying hierarchy can be evaluated explicitly. We introduce a hierarchical token-merging autoencoder that recursively combines adjacent tokens into increasingly abstract representations, compressing entire sequences through a variational information bottleneck. The model jointly learns representations and hierarchical structure by marginalising over possible binary trees using dynamic programming. To address training instability, we combine independently trained models through unsupervised consensus and distillation. Experiments on a synthetic, ambiguous grammar demonstrate the emergence of meaningful hierarchical structure without structural supervision. The resulting model achieves 93% grammatical parse-validity and 93% exact reconstruction on held-out sequences, compared with 42% parse-validity for the strongest trivial baseline. These findings establish a proof of concept for hierarchical token merging as a mechanism for unsupervised structure discovery and abstraction learning.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Hierarchy, Autoencoder, Artificial intelligence, Abstraction, Representation (politics)

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange