Augmenting large language models with external tools could help solve more challenging problems. Current methods for tool learning include supervised methods and in-context learning, but face challenges when scaling to a large number of tools.

Source: https://blog.research.google/

Wearable devices collecting data on physiological and behavioral signals can promote healthy behaviors, detect diseases, and improve treatments. However, raw data is hard to interpret. Algorithms have been developed to convert sensor outputs into more understandable representations. These algorithms typically use supervised, discriminative models designed to detect specific events.

Source: https://blog.research.google/

Cities face traffic congestion challenges, requiring a quantitative understanding of traffic behavior. Congestion functions capture congestion at roadway segment levels, aiding in navigation and traffic flow optimization.

Source: https://blog.research.google/

Augmentative and alternative communication (AAC) devices help individuals with disabilities communicate, but they suffer from slow text-entry speeds. A major issue is the eye fatigue and time cost from numerous keystrokes. One solution is to develop techniques that predict upcoming text to reduce the number of required keystrokes, improving communication quality for these users.

Source: https://blog.research.google/

Understanding population dynamics is crucial for addressing complex social problems, but accurate predictions remain a challenge due to limitations in traditional data sources.

Source: https://blog.research.google/

A new rule proposal uses algorithms to identify 'significant' teams in a sports tournament, reducing potential gains from manipulation. Tournament outcomes are categorized into five groups based on proximity to a team winning all games. If a team wins all games, it's declared the winner. In other cases, a winner is randomly selected or identified manipulation beneficiaries.

Source: https://blog.research.google/

The ionosphere's free electron density affects satellite navigation systems like GPS, causing errors due to seasonal, temporal, and geographical fluctuations. Solar storms can also disrupt satellite communication and navigation.

Source: https://blog.research.google/

Researchers are exploring a unified model to predict both human attention and decision-making behaviors like preferences and ratings.

Source: https://blog.research.google/

A model generates personalized portraits by combining an image of a person's face with a text prompt describing the desired style, using adapters to capture face nuances.

Source: https://blog.research.google/

A new AI model for flood forecasting has been developed, improving predictions in ungauged watersheds with the same accuracy as the previous model at a 7-day lead time.

Source: https://blog.research.google/