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.
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).
totetsu | 4 hours ago
ruther4d | 4 hours ago
totetsu | 3 hours ago
andrewbutts | 58 minutes ago
pishpash | 2 hours ago
rodhruv | 3 hours ago
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).