Second Conference on Language Modeling

The Second Conference on Language Modeling (COLM 2025) is a dedicated forum for research on language models at all scales, from theoretical foundations to the largest neural architectures that have transformed natural language processing. Topics include training dynamics and scaling laws, evaluation methodology, alignment and safety of language models, efficient architectures, emergent capabilities, multilingual and low-resource language modeling, and the societal implications of deploying large generative models.

COLM fills a specific niche in the conference landscape by centering the language model itself — its properties, behaviors, and risks — rather than treating language modeling as a means to other NLP ends. It attracts machine learning researchers, AI safety researchers, computational linguists, and policy-oriented scientists who want a community focused on the fundamental questions raised by the current generation of powerful language models.