The field of large language models (LLMs) is rapidly advancing, with a focus on improving their ability to perform autonomous task-solving in complex environments. Recent developments have seen the introduction of novel architectures and techniques that enable LLMs to better generalize to new situations and adapt to changing conditions. One key area of research is the development of methods for grounding LLMs in their environment, allowing them to understand and interact with the world around them in a more meaningful way. Another area of focus is the improvement of LLMs' ability to reason over long horizons, enabling them to perform tasks that require extended sequences of actions. Notable papers in this area include Grounded Test-Time Adaptation for LLM Agents, which proposes two complementary strategies for adapting LLM agents to novel environments, and From Experience to Strategy: Empowering LLM Agents with Trainable Graph Memory, which introduces a novel graph memory framework for enhancing LLMs' ability to utilize prior experiences in guiding current decisions. Additionally, papers such as Beyond Fact Retrieval: Episodic Memory for RAG with Generative Semantic Workspaces and Procedural Knowledge Improves Agentic LLM Workflows have made significant contributions to the field, demonstrating the potential for LLMs to be used in a wide range of applications, from web navigation to robotic reinforcement learning.