Previous-Token Prediction Based LLM Inversion for Prompt Reconstruction

dalmo3 34 points 6 comments November 19, 2026
arxiv.org · View on Hacker News

Discussion Highlights (2 comments)

totetsu

Imagine if you could take the AI marketing slop, or political smear campaigns, texts, and Images, and see the misanthropic sludge that was used as a prompt to generate it.

rodhruv

I don’t think it’s nearly as impressive as it is hyped to be: They only ever evaluate on 100 prompts, and the error bars on their headline metrics are +-4-5. The headline result (65%) only applies to Qwen3-0.6B chat; for GPT-4o, it’s 11%, and for Llama-2, 17%. It’s fine-tuned on 400 ShareGPT prompts, and I suspect most of the gains are due to learning the ShareGPT prompt format. Indeed, if fine-tuned on a different dataset, it only achieves 11%. The paper also has several misleading claims, like calling it “black-box” (needs the tokenizer) or “data-free” (only during pretraining, but see their own appendix).

Semantic search powered by Rivestack pgvector
13,334 stories · 126,150 chunks indexed