Show HN: ECX a 'Jail-Fix' for RLHF Neutrality Loops in LLMs

I'm a mechanical engineer by trade. I built this ECX logic because I got tired of the 'statistical mean' in LLMs overriding my specific data-point. If anyone is actually running the script from the Zenodo link, I'd love to hear where the 'Inverse Square' logic stalls out for you.