Hidden Signals: Disrupting AI Collusion Through Representational Control
New research demonstrates a defense against covert communication in multi-agent learning by reshaping the internal representations agents use to coordinate.
Science
Beyond Roth-Lempel: Designing Powerful Codes for Quantum Error Correction
This review details the construction of novel Euclidean and Hermitian LCD codes derived from generalized Roth-Lempel codes, offering a new approach to building robust error-correcting systems.
Pricing Catastrophe Risk with Quantum Speedup
![Discretization error overwhelmingly dominates performance at high tail percentiles, evidenced by a [latex]146K[/latex] error at the 90th percentile, while a quantum advantage of 2-2.5× emerges in oracle-model comparisons only at the 95th and 97th percentiles, suggesting a narrow operating regime for quantum benefit.](https://arxiv.org/html/2603.15664v1/plot_exp4_tail.png)
New research demonstrates the potential of quantum amplitude estimation to significantly improve the accuracy and efficiency of tail-risk pricing for catastrophe insurance.
Quantum Chains: A New Take on Collapse
A novel theory proposes that the likelihood of quantum state collapse increases with the interconnectedness of quantum degrees of freedom, potentially resolving the long-standing measurement problem.
Untangling Knots with Quantum Algebra
New research reveals a surprising link between quantum cluster algebras and knot theory, offering a novel approach to understanding the complex properties of these mathematical structures.
Untangling Strong Correlations with Quantum Algorithms
![The study rigorously examines the behavior of exchange-correlation (XC) functionals as a function of density, specifically at an interaction strength of [latex]U=4[/latex], revealing that both the XC energy density [latex]\varepsilon_{XC} = E_{XC}/L[/latex] and the corresponding XC potential [latex]v_{XC}[/latex] remain consistent across varying lattice sizes along the non-magnetic line [latex]n_{\uparrow} = n_{\downarrow}[/latex], while spin-resolved analysis demonstrates comparable trends for XC energy and potential [latex]v_{XC}^{\uparrow}[/latex] as a function of [latex]n_{\uparrow}[/latex], confirmed by agreement between Variational Quantum Eigensolver (VQE) results and exact diagonalization calculations performed at half filling with [latex](N_{\uparrow}, N_{\downarrow}) = (L/2, L/2)[/latex] for even <i>L</i>, utilizing a minimum number of evolution layers to ensure fidelity exceeding 0.99 in each spin sector.](https://arxiv.org/html/2603.16605v1/x4.png)
Researchers are leveraging the power of quantum computing to calculate the complex interactions within strongly correlated materials, paving the way for more accurate material simulations.
Building Trust in a Quantum Future: Secure Multi-Agent Systems
![The system proposes an agentic artificial intelligence architecture fundamentally secured against quantum attacks through its very design, bypassing the need for post-hoc cryptographic defenses often vulnerable to future computation advances [latex] \implies [/latex] a proactive approach to long-term security in increasingly complex AI systems.](https://arxiv.org/html/2603.15668v1/HighLevelArchwithagentblock.png)
A new framework integrates post-quantum cryptography and quantum communication to establish resilient and trustworthy interactions between distributed AI agents.
Capturing Coding Wisdom: How Git History Can Power AI Agents
A new protocol leverages the often-overlooked data within git commit messages to provide AI coding assistants with valuable context and preserve institutional knowledge.
F-Theory’s Limits: When String Theory Breaks Down
New research reveals fundamental obstructions to consistent compactifications of F-theory, showing how quantum effects invalidate standard perturbative expansions.
Beyond Mesons: Uncovering the Secrets of Exotic Tetraquarks
New research delves into the structure and decay of tetraquark particles containing hidden strangeness, potentially revealing new insights into the strong force.