Structured decoding enables large language models (LLMs) to generate outputs in formats required by downstream systems, such as HTML or JSON. However, existing methods suffer from efficiency bottlenecks due to grammar compilation, state tracking, and mask creation. We observe that many real-world tasks embed strong prior knowledge about output structure. Leveraging this, we propose a decomposition of constraints into static and dynamic components -- precompiling static structures offline and instantiating dynamic arguments at runtime using grammar snippets. Instead of relying on pushdown automata, we employ a compositional set of operators to model regular formats, achieving lower transition latency. We introduce wgrammar, a lightweight decoding engine that integrates domain-aware simplification, constraint decomposition, and mask caching, achieving up to 250x speedup over existing systems. wgrammar's source code is publicly available at https://github.com/wrran/wgrammar.

We provide a mathematical argument showing that, given a representation of lexical items as functions (wavelets, for instance) in some function space, it is possible to construct a faithful representation of arbitrary syntactic objects in the same function space. This space can be endowed with a commutative non-associative semiring structure built using the second Renyi entropy. The resulting representation of syntactic objects is compatible with the magma structure. The resulting set of functions is an algebra over an operad, where the operations in the operad model circuits that transform the input wave forms into a combined output that encodes the syntactic structure. The action of Merge on workspaces is faithfully implemented as action on these circuits, through a coproduct and a Hopf algebra Markov chain. The results obtained here provide a constructive argument showing the theoretical possibility of a neurocomputational realization of the core computational structure of syntax. We also present a particular case of this general construction where this type of realization of Merge is implemented as a cross frequency phase synchronization on sinusoidal waves. This also shows that Merge can be expressed in terms of the successor function of a semiring, thus clarifying the well known observation of its similarities with the successor function of arithmetic.

We make the case for language models over logical forms (LFLMs), arguing that such models are more data-efficient than their textual counterparts. To that end, we introduce the Graph-based Formal-Logical Distributional Semantics (GFoLDS) prototype, a pretrained LM over graph representations of logical forms, as a proof-of-concept of LFLMs. Using GFoLDS, we present strong experimental evidence that LFLMs can leverage the built-in, basic linguistic knowledge inherent in such models to immediately begin learning more complex patterns. On downstream tasks, we show that GFoLDS vastly outperforms textual, transformer LMs (BERT) pretrained on the same data, indicating that LFLMs can learn with substantially less data than models over plain text. Furthermore, we show that the performance of this model is likely to scale with additional parameters and pretraining data, suggesting the viability of LFLMs in real-world applications.

Despite tremendous progress in neuroscience, we do not have a compelling narrative for the precise way whereby the spiking of neurons in our brain results in high-level cognitive phenomena such as planning and language. We introduce a simple mathematical formulation of six basic and broadly accepted principles of neuroscience: excitatory neurons, brain areas, random synapses, Hebbian plasticity, local inhibition, and inter-area inhibition. We implement a simulated neuromorphic system based on this formalism, which is capable of basic language acquisition: Starting from a tabula rasa, the system learns, in any language, the semantics of words, their syntactic role (verb versus noun), and the word order of the language, including the ability to generate novel sentences, through the exposure to a modest number of grounded sentences in the same language. We discuss several possible extensions and implications of this result.

Quantitative facts are continually generated by companies and governments, supporting data-driven decision-making. While common facts are structured, many long-tail quantitative facts remain buried in unstructured documents, making them difficult to access. We propose the task of Quantity Retrieval: given a description of a quantitative fact, the system returns the relevant value and supporting evidence. Understanding quantity semantics in context is essential for this task. We introduce a framework based on description parsing that converts text into structured (description, quantity) pairs for effective retrieval. To improve learning, we construct a large paraphrase dataset using weak supervision based on quantity co-occurrence. We evaluate our approach on a large corpus of financial annual reports and a newly annotated quantity description dataset. Our method significantly improves top-1 retrieval accuracy from 30.98 percent to 64.66 percent.

The CoNLL-2013 shared task was devoted to grammatical error correction. In this paper, we give the task definition, present the data sets, and describe the evaluation metric and scorer used in the shared task. We also give an overview of the various approaches adopted by the participating teams, and present the evaluation results.

We present a family of encodings for sequence labeling dependency parsing, based on the concept of hierarchical bracketing. We prove that the existing 4-bit projective encoding belongs to this family, but it is suboptimal in the number of labels used to encode a tree. We derive an optimal hierarchical bracketing, which minimizes the number of symbols used and encodes projective trees using only 12 distinct labels (vs. 16 for the 4-bit encoding). We also extend optimal hierarchical bracketing to support arbitrary non-projectivity in a more compact way than previous encodings. Our new encodings yield competitive accuracy on a diverse set of treebanks.

While tokenization is a key step in language modeling, with effects on model training and performance, it remains unclear how to effectively evaluate tokenizer quality. One proposed dimension of tokenizer quality is the extent to which tokenizers preserve linguistically meaningful subwords, aligning token boundaries with morphological boundaries within a word. We expand MorphScore (Arnett & Bergen, 2025), which previously covered 22 languages, to support a total of 70 languages. The updated MorphScore offers more flexibility in evaluation and addresses some of the limitations of the original version. We then correlate our alignment scores with downstream task performance for five pre-trained languages models on seven tasks, with at least one task in each of the languages in our sample. We find that morphological alignment does not explain very much variance in model performance, suggesting that morphological alignment alone does not measure dimensions of tokenization quality relevant to model performance.

We provide a novel mathematical implementation of tree-adjoining grammars using two combinatorial definitions of graphs. With this lens, we demonstrate that the adjoining operation defines a pre-Lie operation and subsequently forms a Lie algebra. We demonstrate the utility of this perspective by showing how one of our mathematical formulations of TAG captures properties of the TAG system without needing to posit them as additional components of the system, such as null-adjoining constraints and feature TAG.

The DisCoCirc framework for natural language processing allows the construction of compositional models of text, by combining units for individual words together according to the grammatical structure of the text. The compositional nature of a model can give rise to two things: compositional generalisation -- the ability of a model to generalise outside its training distribution by learning compositional rules underpinning the entire data distribution -- and compositional interpretability -- making sense of how the model works by inspecting its modular components in isolation, as well as the processes through which these components are combined. We present these notions in a framework-agnostic way using the language of category theory, and adapt a series of tests for compositional generalisation to this setting. Applying this to the DisCoCirc framework, we consider how well a selection of models can learn to compositionally generalise. We compare both quantum circuit based models, as well as classical neural networks, on a dataset derived from one of the bAbI tasks, extended to test a series of aspects of compositionality. Both architectures score within 5% of one another on the productivity and substitutivity tasks, but differ by at least 10% for the systematicity task, and exhibit different trends on the overgeneralisation tasks. Overall, we find the neural models are more prone to overfitting the Train data. Additionally, we demonstrate how to interpret a compositional model on one of the trained models. By considering how the model components interact with one another, we explain how the model behaves.