Artificial intelligence, machine learning, computer science and mathematics.
A survey of human-AI collaboration argues that more capable AI systems do not automatically produce reliable partnerships. It identifies human-centered design, governance and attention to high-stakes uses as central to making these systems trustworthy and beneficial.
Deep-learning tools for predicting RNA splicing can rely on genomic clues unrelated to splicing and miss important effects of RNA structure, a study finds. These weaknesses were linked to systematic errors, especially when the tools evaluated sequences that differed from those used for training.
AI vulnerability detectors often perform less well on software projects outside their training data. A study finds that cleaner, more varied training data and an encoder-based design improved detection, including a reported 6.8% gain in recall on the BigVul benchmark.
Researchers developed a quantum method for the maximum cut problem and proved a stronger worst-case guarantee than a three-round version of a widely used approach. On IBM quantum devices with 72 and 148 qubits, the method exceeded a known classical hardness threshold, while the comparison method did not do so in the 148-qubit test.
A study finds that shared data from split-learning phishing detectors can reveal private webpage information. A new attack using large language models outperformed conventional reconstruction attacks in experiments on real-world phishing datasets.
A lightweight AI system detected marine organisms in degraded underwater images with lower computing demands than YOLOv8m. Tests on a dataset and an embedded computer suggest it could support faster, more energy-efficient monitoring on autonomous underwater vehicles.
A preprint study found that the same hidden instruction could produce very different results in AI agents depending on whether it arrived as tool information or tool output, and what surrounded it. The findings have not been peer reviewed and may change.
A marker-free augmented-reality system projected planned robotic-surgery port locations onto a human phantom. The results show consistent alignment between the digital plan and the physical body model, but the system has not yet been evaluated in patients.
A simulation of contact tracing in Seoul and Busan found that leaving infected people out of the system had a sharper effect on epidemic spread than missing individual contacts. Contact tracing broke down at an estimated omission rate of about 4% in Seoul and 10% in Busan, suggesting that cities may need different completeness targets.
Researchers developed a light-based AI navigation architecture and tested part of it with a laser array. Simulations reported an 80% ± 8.3% success rate, while the tested laser activations used an estimated 0.78 nJ of energy and took 191.20 picoseconds under ideal parallel operation.
A new tool called plsMD reconstructs antibiotic-resistance plasmids from short-read bacterial genome data. It outperformed existing tools on two datasets, although its performance was lower on a newly sequenced-isolate dataset than on an established benchmark.
A robot completed 91.67% of 240 real-world manipulation trials using an explicit, step-by-step reasoning process. The system did not rely on task-specific demonstrations or manipulation-policy training, but the study does not show how it compares with other systems or performs beyond the designed tests.
A review finds that current safety certification rules do not map cleanly onto the behavior of large language models. It argues that safety-critical systems should certify the surrounding control system, keeping the language model in a role whose outputs are independently checked.
A full-round analysis of the SCAN-C cipher found that its unusual XNOR key mixing adds no security compared with XOR. The researchers also report that the specified 12-round design can be attacked in an analysis requiring around 2^55 chosen plaintexts and 2^69 encryptions.
In a large human, multicenter cohort study, researchers developed and tested an artificial intelligence model for classifying stomach biopsy images. The model showed high reported accuracy in external and prospective testing and improved nine pathologists’ results in an auxiliary experiment, but the study did not establish effects on patient outcomes or performance in every clinical setting.
