Kevin Collins: Perhaps others have seen it already, but I found this pre-print (first posted in September) deeply troubling, raising concerns about how LLMs used for classification tasks in research open new researcher-degrees-of-freedom, which they call "LLM-hacking" (akin to p-hacking) arxiv.org/pdf/2509.08825

Perhaps others have seen it already, but I found this pre-print (first posted in September) deeply troubling, raising concerns about how LLMs used for classification tasks in research open new researcher-degrees-of-freedom, which they call "LLM-hacking" (akin to p-hacking) arxiv.org/pdf/2509.08825

https://arxiv.org/pdf/2509.08825