arXiv subjects

Alberto Cetoli

Publications and source records attributed to Alberto Cetoli.

7 recordsLinked to original sources

arXiv2026

Sleight of Word Benchmark: Can Language Models Notice If Their Own Output Was Tampered With?

The output of a Language Model can be tampered with \emph{while} the model is writing it. A simple test can thus be constructed by evaluating the model's perception of this external perturbation. In this spirit, a simple benchmark is built in which a single word is consistently substituted with another in the generation process. We call this method \emph{Sleight of Word}. Two distinct axes are measured: metrics that relate to the model's surprise, as well as an evaluation of the textual reaction for 19 different open-weight language models.

cs.CL↗

arXiv2021

Dispatcher: A Message-Passing Approach To Language Modelling

This paper proposes a message-passing mechanism to address language modelling. A new layer type is introduced that aims to substitute self-attention for unidirectional sequence generation tasks. The system is shown to be competitive with existing methods: Given N tokens, the computational complexity is O(N logN) and the memory complexity is O(N) under reasonable assumptions. In the end, the Dispatcher layer is seen to achieve comparable perplexity to prior results while being more efficient.

cs.CL↗

arXiv2021

Semantic Reasoning with Differentiable Graph Transformations

This paper introduces a differentiable semantic reasoner, where rules are presented as a relevant set of graph transformations. These rules can be written manually or inferred by a set of facts and goals presented as a training set. While the internal representation uses embeddings in a latent space, each rule can be expressed as a set of predicates conforming to a subset of Description Logic.

cs.AI↗

arXiv2020

Pynsett: A programmable relation extractor

This paper proposes a programmable relation extraction method for the English language by parsing texts into semantic graphs. A person can define rules in plain English that act as matching patterns onto the graph representation. These rules are designed to capture the semantic content of the documents, allowing for flexibility and ad-hoc entities. Relation extraction is a complex task that typically requires sizable training corpora. The method proposed here is ideal for extracting specialized ontologies in a limited collection of documents.

cs.CL↗

arXiv2018

Named Entity Disambiguation using Deep Learning on Graphs

We tackle \ac{NED} by comparing entities in short sentences with \wikidata{} graphs. Creating a context vector from graphs through deep learning is a challenging problem that has never been applied to \ac{NED}. Our main contribution is to present an experimental study of recent neural techniques, as well as a discussion about which graph features are most important for the disambiguation task. In addition, a new dataset (\wikidatadisamb{}) is created to allow a clean and scalable evaluation of \ac{NED} with \wikidata{} entries, and to be used as a reference in future research. In the end our results show that a \ac{Bi-LSTM} encoding of the graph triplets performs best, improving upon the baseline models and scoring an \rm{F1} value of $91.6\%$ on the \wikidatadisamb{} test set

cs.CL↗

arXiv2011

Correlations and superfluidity of a one-dimensional Bose gas in a quasiperiodic potential

We consider the correlations and superfluid properties of a Bose gas in an external potential. Using a Bogoliubov scheme, we obtain expressions for the correlation function and the superfluid density in an arbitrary external potential. These expressions are applied to a one-dimensional system at zero temperature subject to a quasiperiodic modulation. The critical parameters for the Bose glass transition are obtained using two different criteria and the results are compared. The Lifshits glass is seen to be the limiting case for vanishing interactions.

cond-mat.dis-nn↗

arXiv2011

Towards a Bose glass transition in an optical Penrose quasicrystal

We study numerically a 2D Bose-Einstein condensate in a quasiperiodic array of potential peaks, assumed to be generated by superimposing five blue detuned laser beams. By using a Bogoliubov ansatz for the excitations we show that the system approaches a gapless, insulating phase upon increasing the potential, consistent with a Bose glass phase. The characteristics of the transition in terms of phase correlations, oscillatory modes, and superfluid fraction are discussed.

cond-mat.dis-nn↗