arXiv subjects

Gabriel Lefundes

Publications and source records attributed to Gabriel Lefundes.

5 recordsLinked to original sources

arXiv2026

Structure Constants from Q-Systems and Separation of Variables

We introduce a novel method to compute structure constants from Q-functions in the scalar sector of planar N=4 super Yang-Mills (SYM) and related theories. The method derives from operatorial as well as functional separation of variables, and the structure constants are expressed as determinants of matrices whose entries are integrals over products of Q-functions. In this framework, each operator is twisted by an external angle, mirroring the cusped Maldacena-Wilson loop. The structure constants of local single-trace operators in N=4 SYM are recovered in the untwisting limit, where we obtain a one-to-one correspondence between our key building blocks and those of the Hexagon formalism. Retaining appropriate twists, our structure constants also perfectly match those of the orbifold points of N=4 SYM. Our results thus far are valid at leading order in the weak-coupling expansion, but their formulation in terms of Q-functions provides a natural starting point for including loop corrections. Many of the methods we develop in this work apply more generally to the computation of correlation functions in integrable models.

hep-th↗

arXiv2025

Exact Three-Point Functions in $\mathcal{N}=2$ Superconformal Field Theories: Integrability vs. Localization

We propose an integrability approach for planar three-point functions at finite coupling in $\mathcal{N}=2$ superconformal field theories obtained as $\mathbb{Z}_K$ orbifolds of $\mathcal{N}=4$ super Yang-Mills (SYM). Generalizing the hexagon formalism for $\mathcal{N}=4$ SYM, we reproduce the structure constants of Coulomb branch operators, previously obtained by supersymmetric localization, as exact functions of the 't Hooft coupling. Our analysis explains the common physical origin of Fredholm kernels in integrability and localization, and hints at structures after the resummation in the hexagon formalism.

hep-th↗

arXiv2023

Hexagons and the classical limit

In planar maximally supersymmetric Yang-Mills, we can compute three-point functions at weak coupling using the so-called hexagonalization formalism. The main objects in this framework are called hexagons. We are interested in two sectors of the theory, spanned by operators built entirely from scalar fields, and by spinning operators, respectively. By reviewing the analytic properties of these building blocks, we find new representations for them at weak coupling where the two sectors are on equal footing. We compute the classical limit of the hexagons and of correlation functions in both sectors for some choices of polarizations and our results match previous predictions in the literature.

hep-th↗

arXiv2021

Attention-based fusion of semantic boundary and non-boundary information to improve semantic segmentation

This paper introduces a method for image semantic segmentation grounded on a novel fusion scheme, which takes place inside a deep convolutional neural network. The main goal of our proposal is to explore object boundary information to improve the overall segmentation performance. Unlike previous works that combine boundary and segmentation features, or those that use boundary information to regularize semantic segmentation, we instead propose a novel approach that embodies boundary information onto segmentation. For that, our semantic segmentation method uses two streams, which are combined through an attention gate, forming an end-to-end Y-model. To the best of our knowledge, ours is the first work to show that boundary detection can improve semantic segmentation when fused through a semantic fusion gate (attention model). We performed an extensive evaluation of our method over public data sets. We found competitive results on all data sets after comparing our proposed model with other twelve state-of-the-art segmenters, considering the same training conditions. Our proposed model achieved the best mIoU on the CityScapes, CamVid, and Pascal Context data sets, and the second best on Mapillary Vistas.

cs.CV↗

arXiv2020

On estimating gaze by self-attention augmented convolutions

Estimation of 3D gaze is highly relevant to multiple fields, including but not limited to interactive systems, specialized human-computer interfaces, and behavioral research. Although recently deep learning methods have boosted the accuracy of appearance-based gaze estimation, there is still room for improvement in the network architectures for this particular task. Therefore we propose here a novel network architecture grounded on self-attention augmented convolutions to improve the quality of the learned features during the training of a shallower residual network. The rationale is that self-attention mechanism can help outperform deeper architectures by learning dependencies between distant regions in full-face images. This mechanism can also create better and more spatially-aware feature representations derived from the face and eye images before gaze regression. We dubbed our framework ARes-gaze, which explores our Attention-augmented ResNet (ARes-14) as twin convolutional backbones. In our experiments, results showed a decrease of the average angular error by 2.38% when compared to state-of-the-art methods on the MPIIFaceGaze data set, and a second-place on the EyeDiap data set. It is noteworthy that our proposed framework was the only one to reach high accuracy simultaneously on both data sets.

cs.CV↗