Institution profile

University of Barcelona

Academic institutioneurope · es
Official website
Research library198linked papers
Opportunities0open roles
Selected work

Representative Papers

Measure-to-measure interpolation using Transformers

Nov 07, 2024arXiv.org

This paper investigates the expressive power of Transformers as arbitrary input-to-output measure mappings. Method: We reformulate Transformers from a measure-theoretic perspective, modeling them as differentiable maps on continuous measure spaces—departing from conventional discrete token-based interpretations. Leveraging continuity equations to describe particle dynamics, we design an attention mechanism incorporating spherical geometry constraints and optimal transport theory. Contribution/Results: We propose the first Transformer architecture provably capable of exact matching between arbitrary input–target measure pairs. Under the minimal assumption that a transport map exists between each pair, a single model achieves precise matching for N arbitrary measure pairs. We establish theoretical completeness by proving that Transformers serve as universal interpolators between measures and provide explicit parameter constructions. This work fundamentally characterizes the expressive capacity of Transformers for measure transformation tasks.

10 citations1 influentialRead paper

Degree-Preserving Gödel Logics with an Involution: Intermediate Logics and (Ideal) Paraconsistency

Jan 13, 2026Outstanding Contributions to Logic

This study investigates intermediate logics situated between degree-preserving Gödel fuzzy logic with involution and classical propositional logic, focusing on their paraconsistent behavior relative to involutive negation in the finite-valued setting. By introducing “saturated paraconsistency”—a notion strictly weaker than ideal paraconsistency—and combining tools from algebraic logic, the classification of intermediate logics, and finite-valued fuzzy logic, the work fully characterizes the boundaries of all ideal and saturated paraconsistent logics lying between the n-valued Gödel involutive logic and classical logic. Furthermore, it identifies a broad class of saturated paraconsistent logics within intermediate systems of finite-valued Łukasiewicz logic.

6 citations1 influentialRead paper

Warm Starts, Cold States: Exploiting Adiabaticity for Variational Ground-States

Feb 05, 2026

This work addresses the challenge that variational quantum eigensolvers (VQE) often converge to local minima or suffer from barren plateaus in complex energy landscapes, hindering reliable preparation of many-body ground states. To overcome this, the authors propose an iterative strategy inspired by adiabatic evolution, which constructs a discretized deformation path of the Hamiltonian and tracks the ground-state manifold across a sequence of intermediate problems to guide VQE toward the target ground state. The method provides theoretical guarantees on trainability by avoiding regions where the spectral gap closes, thereby significantly enhancing the robustness and scalability of ground-state preparation. Numerical experiments demonstrate that the approach achieves stable convergence even in the presence of measurement noise, effectively circumventing the optimization pitfalls commonly encountered in conventional VQE implementations.

1 citationsRead paper

Tensor-DTI: Enhancing Biomolecular Interaction Prediction with Contrastive Embedding Learning

Jan 09, 2026arXiv.org

This work addresses the limited expressiveness and generalizability of existing drug–target interaction (DTI) prediction models that rely on single-modality representations. The authors propose Tensor-DTI, a novel framework that, for the first time, integrates multimodal embeddings—comprising molecular graphs, protein language models, and binding site predictions—through contrastive learning. A siamese dual-encoder architecture is employed to model chemico-structural interaction features. Tensor-DTI significantly improves performance in distinguishing interacting from non-interacting pairs, outperforming state-of-the-art methods across multiple DTI benchmarks. It yields plausible hit distributions in billion-scale virtual screening against CDK2 and substantially reduces the cost of discovering high-affinity ligands for out-of-family targets. Moreover, the model demonstrates strong interpretability and cross-task extensibility, as evidenced by its applicability to protein–RNA and peptide–protein interactions.

1 citationsRead paper
Recent publications

Latest Papers

Dynamic Polyhedral Logic

Aug 26, 2026

本文通过引入基于多面体语义和路径空间可达性算子的时空多面体可达性逻辑，解决了动态拓扑逻辑中的问题，并证明了其在可逆动力系统中的健全性和完备性。

0 citationsRead paper