Launch note
Read the paper. Then play with it.
A paper can describe a mechanism perfectly and still leave the reader wondering what moves when one assumption changes. Flash Papers turns that mechanism into something you can manipulate.
A companion, not another summary
Summaries answer, “What does this paper say?” A useful lab asks, “What changes when I move this control?” Flash Papers turns an arXiv paper into a single-page explainer with sliders, diagrams, and mini-demos, then gives you a link you can share.
Mechanisms you can test now
LoRA
Set rank to 4 with Q/V, then rank to 2 with Q/K/V/O. Watch the adapter count stay at 65,536.
How a Flash Paper is made
You paste an arXiv or ChinaRxiv link and select Gemini, Claude, or a model through OpenRouter. Your chosen model reads the paper and writes the interactive page. The result opens in a sandboxed viewer. You can keep it private to your account, or share it as a public or unlisted paper.
The source remains authoritative
Every Flash Paper is model-generated. It may omit a condition, overstate a result, or represent a mechanism incorrectly. Check the explainer against the linked source before relying on it for research, teaching, or engineering decisions. Flash Papers helps you interrogate a paper; it does not replace reading one.
Start with a public example, then make your own.