The essence of aligning large LLM models is basically to conform to the existing social order. ChatGPT respects China's order, while China's open-source models, although reluctantly, also tolerate the order in the US. Maybe we can build a revisionist SFT dataset.
Any conversation
I've been enjoying toast to help click through a project. I still mostly use Zed to handwrite code, but it's nice to open a pane and fire up toast from the terminal. paradise-runner.github.io/toast/




