Letting AI Explore Graphs: A New Approach to Reasoning

Researchers are leveraging the power of artificial intelligence to autonomously navigate and learn from complex graph-structured data, unlocking improved performance in reasoning tasks.

Researchers are leveraging the power of artificial intelligence to autonomously navigate and learn from complex graph-structured data, unlocking improved performance in reasoning tasks.

A new analysis reveals that designing effective markets for autonomous agents demands more than simply mirroring human economic principles.

A new dataset and comparative analysis reveals the difficulties of accurately forecasting agricultural commodity prices in a dynamic, developing market.
Researchers have created a unique resource to help artificial intelligence better understand the nuances of financial language and reasoning, moving beyond simple sentiment to capture true market understanding.

A new benchmark reveals the challenges large language models face when applied to complex, real-world financial modeling tasks requiring extended reasoning.
![A novel detection framework leverages high-quality, synthetically generated videos-created by refining real-world captions into prompts for advanced text-to-video generation-and employs a [latex]3D[/latex] patchification approach with the Qwen2.5-VL Vision Transformer to preserve spatiotemporal fidelity and enable robust detection of AI-generated videos, even with variable resolutions and lengths, without the artifacts introduced by conventional downsampling.](https://arxiv.org/html/2604.04634v1/x2.png)
Researchers are developing more effective methods to identify AI-generated videos by focusing on the subtle artifacts left behind in the forgery process.
A new study reveals that the biggest obstacle to AI replicating human-level economics research isn’t technical skill, but the ability to formulate original ideas.

A new approach leverages neural networks to optimize battery energy storage strategies, enhancing implicit balancing and maximizing revenue in dynamic power systems.

Researchers have developed a new automated framework that significantly improves the effectiveness of attacks designed to reveal whether a specific data point was used to train a large vision-language model.

A new framework uses artificial intelligence to dynamically tailor insurance questionnaires, promising more accurate risk profiles and a better user experience.