PaperRobot

PaperRobot is a research artifact, not a tool you'd use daily. The Turing test numbers are the headline: up to 30% of biomedical experts chose its abstracts over human-written ones. That's notable for a 2019 system, but it also means 70% preferred the human version. The incremental pipeline—title to abstract, abstract to conclusion and future work, future work to a new title—is an elegant demonstration of how knowledge graphs and memory-attention networks can structure scientific writing. Pick this if you're an NLP researcher studying knowledge-graph-based generation or automated drafting. The code is open-source, so you can dissect the architecture and extend it. It's also handy for graduate students exploring baseline systems. But don't expect a plug-and-play experience: there's no hosted demo, no API, and you'll need deep expertise to run the code and interpret outputs. Where it bites: the project is dormant since 2019. No updates mean you'll rely on your own maintenance. The original domain was biomedical, so generalizing to other fields requires retraining. And the Turing test preference, while real, doesn't translate to a ready-to-use academic assistant. You'd still need to heavily edit whatever it drafts. Compared to commercial tools like Writefull or Scholarcy, PaperRobot is the opposite: free, but research-only. Writefull gives you polished grammar and style suggestions; Scholarcy summarizes papers. Neither generates entire sections or proposes follow-on titles. If you want that capability today, you're better off with modern LLM-based writing aids—but for understanding the mechanics of automated scientific writing, PaperRobot remains a useful reference.