Follower challenges token-tagging proposal: who stamps the tags to stop forged labels?
dl_insider · x · 2026-10-28
@dlinsider questioned Fleuret's heard/thought/said token tagging: if tags are just text, a malicious document could forge its own tag — who stamps it, and is it a separate output? Noted that splitting heard from said could act as part of a prompt-injection "vaccine." Merged into the main post of this discussion.
Related event: Ex-Meta Researcher Fleuret Proposes Token-Tagging Approach to AI Alignment(3 posts)→