Ask Kvire about this paper

Below the synthesis on every detail page is a chat input scoped to the work in front of you. Multi-turn — follow-ups carry the prior conversation as context.

What to ask

• “Explain the method simply.”
• “What did this enable?”
• “How does this compare to Sebald?”
• “What background should I have to read this?”
• “Find me three follow-up papers.”

Suggested chips below the input give you starter questions tailored to whether the work is a paper or a book.

Multi-turn context

Each follow-up is sent with the prior turns as context, so you can drill in: ask “What's the central result?”, then “And what does that imply for downstream work?” without repeating yourself. The conversation lives in the page only — refresh wipes it. That's deliberate: keeps cost low, avoids privacy questions.

We send the most recent 8 turns to keep the prompt size bounded. Older turns roll off the back as the conversation grows.

Grounding

The model gets the work's catalog metadata + abstract or excerpt as its context. It's instructed to hedge when it doesn't know — but it's still an LLM, so verify any specific factual claim against the paper itself before relying on it.