The field of machine learning is witnessing a significant shift towards multimodal and autonomous systems. Researchers are focusing on developing generalist agents that can interact with computers in a multimodal manner, encompassing text, images, audio, and video. These agents are being designed to integrate tool-based and pure vision agents within a highly modular architecture, enabling them to collaboratively solve decoupled tasks in a step-by-step manner. Additionally, there is a growing interest in automating machine learning workflows, with a focus on developing multi-agent systems that can handle diverse data modalities with minimal human intervention. These systems are being powered by large language models and are demonstrating superior performance on various benchmarks. Furthermore, researchers are exploring the use of metacognitive learning to empower large language models with the ability to reason, reflect, and create, thereby enhancing their ability to perform robotic tasks with minimal demonstrations. Noteworthy papers include:

• InfantAgent-Next, which introduces a generalist agent capable of interacting with computers in a multimodal manner, achieving state-of-the-art results on various benchmarks.
• MLZero, which presents a novel multi-agent framework powered by large language models that enables end-to-end machine learning automation across diverse data modalities.
• R&D-Agent, which introduces a dual-agent framework for iterative exploration, demonstrating its potential to accelerate innovation and improve precision across diverse data science applications.