New research establishes a definitive lower bound for solving a fundamental class of logical problems, even with simplified constraints, and offers algorithmic improvements for a common subproblem.
![A concentration-adjusted slippage analysis across 184 tokens reveals a pronounced liquidity risk within the most extreme 5% of observations, demarcated by a [latex]\mathrm{SaR}^{\mathrm{adj}}(0.95)=3.47\%[/latex] threshold, suggesting heightened vulnerability in those specific assets.](https://arxiv.org/html/2603.09164v1/figure1_slippage_distribution.png)
A novel framework, Slippage-at-Risk (SaR), offers a proactive way to assess execution risk in the fast-moving world of perpetual futures trading.
New research reveals a critical trade-off in neural audio codecs: deeper compression can improve speech recognition, but also opens the door to adversarial manipulation.
As quantum software grows in complexity, researchers have developed an automated pipeline to identify and diagnose the root causes of unreliable, or ‘flaky’, tests.
Researchers have established a surprising connection between quantum error-correcting codes and a class of classically-inspired codes used in secure multi-party computation.
Current methods for evaluating automated vulnerability repair tools are significantly overestimating their success rates, masking critical flaws in patched code.
![Spin textures-specifically Rashba, persistent spin helix, and Dresselhaus states-exhibit a striking enhancement of all quantum metric components [latex]g_{\mu\nu}[/latex] when Rashba and Dresselhaus strengths equalize, signaling a unique signature of the persistent spin helix state and demonstrating how geometric properties emerge from spin-orbit interactions.](https://arxiv.org/html/2603.08009v1/x1.png)
Researchers have discovered a way to sensitively detect and characterize a persistent spin helix in materials by leveraging the quantum metric, a geometric property of quantum systems.
This review explores techniques for building robust analog computing systems by leveraging error-correcting codes that maintain signal integrity with remarkably low overhead.
Researchers have developed a novel method for dramatically reducing the size of large language models without sacrificing accuracy, paving the way for faster and more accessible AI.
A new hardware-software co-design eliminates the need for a runtime software trusted computing base, offering a fundamentally more secure approach to embedded system protection.