LitPivot Contextualizes Research Ideas Within a Dynamic Literature Landscape

Ai2 researchers and collaborators from the University of Pennsylvania have developed LitPivot, a system designed to help early-stage researchers reshape research ideas by contextualizing them against relevant scientific literature. LitPivot addresses what the team calls a literature-initiated pivot, when reading prior work reveals that an idea has already been addressed, exposes an unresolved problem, or suggests a new direction of inquiry. LitPivot combines literature retrieval through Asta Paper Finder with proposal editing support. Users write a short proposal and enable facet highlighting, which divides the idea into its problem, contribution and evaluation facets. When a facet is selected, the system groups related papers around that facet, ranks them by relevance and lets the researcher choose which studies to use for assessment. LitPivot then generates critiques supported by excerpts from selected papers and suggests targeted revisions, which users can accept, edit or reject. As the proposal changes, the system reranks literature and flags passages that may no longer be coherent, such as an evaluation plan that no longer fits a revised contribution. In a user study with 17 researchers, domain experts rated proposals created with LitPivot as better grounded in literature than those developed with a text editor and an AI chat tool. Participants also reported stronger understanding of relevant research and perceived their revised ideas as more novel and useful. A follow-up study with five researchers using LitPivot on ongoing projects showed the tool supported both early idea exploration and refinement of more mature proposals, including framing contributions in new ways. LitPivot will be presented at UIST 2026 in Detroit. The work fits Ai2’s broader effort to develop AI tools for scientific discovery, including hypothesis generation, literature synthesis and evidence-based idea development. By making the choice of supporting literature part of the revision process, LitPivot aims to help researchers connect evolving proposals to existing knowledge while retaining control over both the draft and the evidence used to evaluate it.