/discover

To find research papers to read next without downloading, use a discovery tool that generates a ranked shortlist of candidate papers from anchor, topic, wiki, or venue/year seeds. This proposal-only approach strictly separates recommendations from actual ingestion, ensuring no auto-ingestion occurs.

Semantic scholar paper discovery is the process of generating a relevance-ranked shortlist of candidate papers. It ranks recommendations by routing specific seed modes—such as anchor, topic, wiki, or venue/year—into the discovery tool and presenting rationales for each proposed paper.

Yes, you can get paper recommendations based on a specific conference or anchor paper. The discovery workflow supports multiple seed modes, allowing you to drive candidate generation from specific anchor papers, keywords, or venue/year conference lists to produce a tailored shortlist.

To avoid repeated paper recommendations in your research workflow, apply local wiki dedup. This mechanism filters out already-ingested papers from the ranked shortlist by checking against the wiki-root, ensuring only new candidate papers are proposed for your reading list.

No, the paper discovery skill does not automatically save full text to your wiki. It enforces strict separation between proposal and action by never writing paper content to wiki or raw formats, only logging a run for certain seed modes to maintain safe boundaries.