Comment by teleforce

It's interesting to note that the paper benchmarked its chess playing performance against GPT-3.5-turbo-instruct, the only well performant LLM in the posted article.