Shallow Preference Signals: Large Language model aligns even better without truncated data?

Best AI papers explained - Shallow Preference Signals: Large Language model aligns even better without truncated data?