The work is scheduled for presentation at UIST 2026 in Detroit. LitPivot uses papers retrieved through Asta Paper Finder.
A researcher writes a short proposal, then selects a highlighted facet: problem, proposed contribution, or evaluation plan. LitPivot groups and reranks papers for that facet, lets the researcher choose evidence, and returns a critique, supporting excerpts, and a proposed revision. Users can accept, edit, or reject revisions; the system can flag other proposal passages that may no longer fit after a change.
Ai2 reports a lab study with 17 researchers. Each used LitPivot for one assigned idea and a text editor plus an AI tool for another, with the same paper collection in both conditions. Blinded domain experts rated LitPivot proposals as better grounded in relevant literature. Participants also reported stronger understanding and higher perceived novelty and usefulness. A follow-up involved five researchers using the system on ongoing projects in 30-minute sessions.
The source does not give effect sizes, describe a public release, state pricing or access terms, or establish that LitPivot improves the quality, novelty, or eventual outcomes of research beyond these studies.