New research explores how analyzing internal model states can effectively detect malicious prompts targeting large language models, offering a practical defense against emerging cybersecurity risks.
The ARC Prize 2025 technical report details significant progress toward artificial general intelligence, revealing how systems are learning to improve themselves through iterative refinement.
A new active learning framework boosts outlier detection by first mastering the characteristics of normal data.
A new study demonstrates the potential of artificial intelligence to detect depression by analyzing electroencephalography (EEG) data, offering a promising avenue for objective diagnosis.
Researchers have released a comprehensive benchmark to help detect increasingly realistic videos created by artificial intelligence.
![The self-reflection framework reveals a performance decline correlated with increasing levels of spurious correlation-specifically, as the correlation coefficient [latex]b[/latex] rises from 0.5 to 0.9, the system’s efficacy diminishes, demonstrating the framework’s sensitivity to deceptive patterns within the data.](https://arxiv.org/html/2601.11021v1/x3.png)
A new self-reflection framework helps graph neural networks identify and eliminate misleading correlations, leading to more reliable and consistent explanations.
A new approach combines automated text analysis with human expertise to make sense of complex stories hidden within large collections of text.
The mere presence of advanced artificial intelligence, even if unused, can create incentives for strategic manipulation of regulatory systems and ultimately shift market dynamics.
Researchers are drawing inspiration from the human hippocampus to build more effective systems for identifying fraudulent activity in online finance.
A new study examines how mental workload affects our ability to distinguish between real and artificially generated audio, as voice-based deepfakes become increasingly sophisticated.