Review queue

Entities that need attention. None of these are gated from view — /review is a working queue, not a filter. Sections: pending-review residue (legacy gating to be retired), orphans with no edges, restate-clusters that may be LLM paraphrase artifacts, alias conflicts (same DOI on two entities).

Legacy entities still at status=pending-review. The loader now writes active by default; promote_pending_review.py runs nightly. These remaining rows are either low-quality (few edges) or genuinely deserve manual review. Expected to trend toward 0.

  • gap
    192
  • thinker
    15

Orphans (60)

≥7 days old, zero edges

Entities created > 7 days ago that never accumulated any relations or citations. Often one-shot LLM hallucinations from extraction. Consider archiving in bulk.

Large restate clusters (9)

≥3 entities mutually-restating

Groups of sentence-shape entities tied by cosine ≥0.90 paraphrase edges. Within-paper clusters have been auto-merged by merge_claims_under_same_paper; what remains is cross-paper restatement — possibly legitimate (multiple papers making the same claim) or possibly a sign the dedup needs retuning.

Alias conflicts (0)

same DOI / arXiv on >1 entity

A single DOI/arXiv/openalex id pointing to multiple entity rows. merge_paper_dupes should have caught these — anything here is either a new dupe since last merge run, or an edge case the script missed.

No alias conflicts. ✓