Smarter Inference: Adapting Large Language Models to Run Faster
A new framework intelligently adjusts how large language models process information, delivering significant performance gains without sacrificing accuracy.
Science
Greening Our Cities: A Blueprint for Sustainable Urban Environments
Asphalt Green is pioneering innovative approaches to urban sustainability, integrating green infrastructure with community engagement to address critical challenges like stormwater management and biodiversity loss.
Cracked Shield: How Optimized AI Security Can Backfire
New research reveals that performance optimizations within secure AI systems can inadvertently expose large language models to full recovery and integrity bypass.
Superconducting Logic Gets a Boost from Error Correction
Researchers demonstrate a practical error-correction encoder for superconducting circuits, enhancing reliability in the face of manufacturing imperfections and faults.
Beyond Single Keys: Scaling Secure Signatures with Trusted Hardware
![The study demonstrates a method for signature generation that directly impacts True Positive Rate (TPR) performance, establishing a quantifiable relationship between the generated signature’s characteristics and the resulting [latex] TPR [/latex] metric.](https://arxiv.org/html/2602.09707v1/figures/TPM_performance.png)
A new framework streamlines multi-party cryptographic operations by distributing the signing process across multiple Trusted Platform Modules.
Beyond On-Chain Data: How Bitcoin Oracles Are Evolving
A new review charts the shift in Bitcoin oracle designs away from traditional on-chain feeds towards privacy-focused, off-chain attestation methods like Discreet Log Contracts.
Quantum Risk Analysis: A Faster Path to Structural Safety
A new quantum algorithm dramatically accelerates the estimation of critical failure probabilities in complex engineering systems, offering a significant advantage over traditional methods.
AI Learns to Find Software Bugs with Human-Like Reasoning
A new framework combines the power of artificial intelligence with formal verification techniques to autonomously discover vulnerabilities in software code.
Uncovering Quantum Code Errors with Intelligent Program Variation
A new testing framework, QEMI, automatically identifies bugs in quantum software by creating and comparing simplified versions of complex programs.
Beyond SPHINCS+: A New Signature Scheme Hardens Post-Quantum Security
![The SpinelHash function pipeline anticipates inevitable decay by constructing a probabilistic structure where each layer’s output-derived from [latex]n[/latex] independent hash computations-serves not as a definitive result, but as the seed for the next, accepting that entropy will ultimately dominate any attempt at perfect preservation.](https://arxiv.org/html/2602.09882v1/x2.png)
Researchers have developed Spinel, a novel signature scheme building on the widely-adopted SPHINCS+ framework, to enhance its theoretical underpinnings and resilience against future attacks.