Comment by abalaji

My guess is that this is a post training (rlhf) artifact on world model prompts. There were likely many “logical inconsistency” prompts which humans coerced to the above response.