The rapid advancements in natural language processing (NLP) within the past few years have sparked a growing interest in the integration of language into robot learning. This has resulted in a growing body of research at the intersection of NLP, computer vision and robotics, covering a wide range of topics, such as human-robot communication, language-driven representation learning, specification of rewards, tasks, and constraints, as well as the utilization of large pretrained language models for control. The first edition of this workshop featured over 100 participants engaging with 9 diverse invited speakers, 2 panel discussions, 14 accepted posters, and 1 live interactive robotics demonstration. 

In last year’s exploratory first edition of the workshop, speakers and participants agreed that the most interesting future challenges are in how language in robot learning can progress towards grounding more modalities. This excitement was reflected in the broader community as well, with the proliferation of LLM and VLM driven robotics research in recent months. Building on top of these ideas, the second edition of this workshop aims to focus on how language can act as a common grounding for multimodal data, across robot embodiments, and beyond offline training methods.