Quantum papers — 2026-09-23

Today's work focuses on building secure systems where parties do not trust each other, specifically looking at spatiotemporal multi-party computation. This approach allows for computing things while keeping sensitive location and time data private, ensuring the results match what actually happened in the real world. The research aims to figure out how to extract this spatiotemporal information from a successful computation and then prove its physical validity using an auxiliary verification protocol.

A key part of this involves defining universal composability for these spatiotemporal MPC protocols. This definition captures privacy, physical consistency, and composability all at once. Researchers have constructed UC-secure commit-and-prove protocols for spatiotemporal knowledge in two settings: the CRS model under LWE against quantum provers without pre-shared entanglement, and the QROM against quantum provers with unbounded pre-shared entanglement. These constructions enable obtaining UC-secure spatiotemporal MPC from semi-honest post-quantum MPC.

Building on this foundation, the framework is being extended to handle quantum functionalities even when the input data is classical in nature. This connection relates directly to other research exploring how learned representations influence the exchange between quantum computing and machine learning, such as using Neural Quantum Embedding to improve classification on noisy hardware.

Today's papers

End-to-End Quantum Semantic Communication with Variational Quantum Neural Networks This paper presents a framework for sending semantic information using quantum machine learning and communication channels. [paper]
Encrypted Redundancy as a Diagnostic Resource: Relational Diagnosis in Quantum Encrypted Cloning This work shows that redundant parts of an encrypted quantum state can be used to diagnose faults without revealing the secret key. [paper]
When Quantum Meets AI: Quantum Methods for Machine Learning and Machine Learning Methods for Quantum Systems This thesis explores how quantum methods can improve machine learning and how machine learning methods can be applied to quantum systems. [paper]
From IceCube to IT-Sphere: A Hybrid Quantum-Classical GNN for Banking IT Root Cause Analysis This paper proposes a hybrid quantum graph neural network approach for analyzing root causes in banking IT operations. [paper]
Bridge of 's: Quantum Circuit Optimization with Schr"odinger Bridges This paper introduces a generative model based on Schrödinger bridges to learn optimized quantum circuits directly from examples. [paper]
Hyperbolic Restricted Boltzmann Machine Neural Quantum State This work constructs a new type of neural quantum state using hyperbolic geometry that better represents complex quantum systems. [paper]
When are bosonic Gaussian states classical to learn? This paper investigates the conditions under which bosonic Gaussian states become classical and can be learned efficiently. [paper]
I Prove, Therefore I Am: Spatiotemporal Multi-Party Computation This paper defines a new security framework for multi-party computation that incorporates physical location and time constraints. [paper]

The papers

Important terms

Spatiotemporal Multi-Party Computation: This is a method for parties who don't trust each other to compute things while keeping sensitive location and time data private. It ensures the final results are physically consistent with real-world events.
Universal Composability (UC): This is a key definition that combines privacy, physical consistency, and composability into one concept for these protocols. It means the system is robust across different scenarios.
Commit-and-Prove Protocols: These are specific types of secure protocols where parties commit to information and then prove later that their actions were correct. This is used here for spatiotemporal knowledge.
LWE and QROM: These are mathematical problems used as the underlying security assumptions for the protocols. LWE relates to quantum provers without pre-shared entanglement, while QROM involves unbounded pre-shared entanglement.
Neural Quantum Embedding: This concept connects learned representations from machine learning with quantum computing. It helps improve classification accuracy even when using noisy hardware.