When AI Gets the Basics Wrong: The Limits of Language Model Reliability
)))))[/latex] as a valid sequence.](https://arxiv.org/html/2601.15714v1/imgs/chatgpt.png)
New research reveals that even the most advanced language models struggle with fundamental tasks, highlighting a critical need for more robust evaluation metrics.
Science
Dark Forces, Chiral Walls: Exploring a New QCD Transition
![The system explores domain-wall profiles within a [latex]Z(3)[/latex] symmetry, demonstrating how chiral backreaction influences their structure and evolution as a natural consequence of the system’s inherent decay.](https://arxiv.org/html/2601.15342v1/Fig4_PRL_pretty.png)
A new theoretical model details how the interplay of confinement and chiral symmetry breaking shapes phase transitions within a hidden sector governed by dark QCD.
Catching Bugs in the Loop: AI-Powered Code Analysis for Python
A new framework leverages the power of locally-run artificial intelligence to identify security vulnerabilities hidden within Python code’s looping structures.
Entangled Hyperons Challenge Local Realism
New analysis of hyperon-antihyperon pairs provides stringent tests of quantum mechanics against hidden-variable theories.
The Detector Delusion: Why AI-Generated Text Is Still Undetectable
A new analysis reveals that current methods for identifying AI-written content are fundamentally flawed and easily bypassed.
Poisoning the Well: How Attackers Can Subvert Retrieval-Augmented Generation
A new study reveals a potent attack framework capable of extracting sensitive data from Retrieval-Augmented Generation systems by intelligently navigating their underlying knowledge sources.
Quantum Chains: Untangling Interactions in Rydberg Atom Systems
![The study reveals that subtle deviations from a simplified [latex]\hat{H}_{PXQ}[/latex] Hamiltonian-attributable to longer-range interactions-manifest as measurable changes in the average classical number of excitations, leakage from the single dimer subspace, and the averaged Rydberg population difference, despite nearly identical dynamics predicted by both the simplified and full [latex]\hat{H}_{0}[/latex] Hamiltonians across a system of length ten.](https://arxiv.org/html/2601.15866v1/fig/Fig6_Ncl_single_excited_v1.png)
New research reveals how strongly interacting chains of Rydberg atoms can exhibit surprisingly predictable dynamics under certain conditions.
Faster Homomorphic Encryption: Multiplying Ciphertexts with Greater Efficiency
New optimizations to ciphertext multiplication are unlocking faster and more practical computations within the realm of homomorphic encryption.
Erasing false Memories from AI: A New Approach to Fighting Misinformation
As large language models become increasingly powerful, ensuring they can ‘unlearn’ incorrect information is critical, but existing techniques falter when applied to compressed AI models.
Unraveling Quantum Fields: Simulating String Breaking with Tensor Networks
Researchers are leveraging tensor networks and qudit-based quantum circuits to explore the complex dynamics of quantum electrodynamics in a simplified 2+1D model.