*> because it's more likely to literally interpret that sequence and go into the... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		airstrike 63 days ago \| parent \| context \| favorite \| on: Our LLM-controlled office robot can't pass butter > because it's more likely to literally interpret that sequence and go into the latent space of drama This always makes me wonder if saying some seemingly random of tokens would make the model better at some other task petrichor fliegengitter azúcar Einstein mare könyv vantablack добро حلم syncretic まつり nyumba fjäril parrot I think I'll start every chat with that combo and see if it makes any difference

yunohn 63 days ago | [–]

There’s actually research being done in this space that you might find interesting: “attention sinks” https://arxiv.org/abs/2503.08908

arjvik 63 days ago | [–]

No Free Lunch theorem applies here!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact