Dear Editors, We are pleased to submit our manuscript entitled "GRAIL: Autonomous Concept Grounding for Neuro-Symbolic Reinforcement Learning" for consideration in the Neurosymbolic Artificial Intelligence journal. This work introduces GRAIL, a framework that grounds relational concepts in neuro-symbolic reinforcement learning agents through environment interaction and weak supervision from large language models. We evaluate on three Atari environments (Kangaroo, Seaquest, and Skiing), demonstrating that learned concepts match or exceed hand-crafted ones. We believe this work is well-suited for this journal as it directly addresses the symbol grounding problem at the intersection of neural and symbolic AI. By removing the need for expert-defined relational semantics, GRAIL paves the way for neuro-symbolic agents that can be deployed in new environments without manual concept engineering. The manuscript has not been published or submitted elsewhere. All authors have approved the submission. Sincerely, Hikaru Shindo, on behalf of all authors