[Exchange] Enrich next 20 hypotheses with PubMed abstracts in evidence
ID: facbd5d5-d55
Priority: 92
Type: one_shot
Status: open
Goal
Continuing from top 3, fetch PubMed abstracts for evidence citations of the next 20 hypotheses by score. 129/149 still lack abstracts.
Acceptance Criteria
☐ Concrete deliverables created
☐ Work log updated with timestamped entry
Work Log
2026-04-15 21:08 PT — Slot 0
- Confirmed task target: hypotheses 41-60 by composite score (next 20, offset 40)
- Discovered 50 unique PMIDs from evidence_for fields lacking abstracts in papers table
- Ran enrich_next20_hypotheses.py: fetched 45/50 PubMed abstracts in one batch
- Stored/updated 45 papers with abstracts (5 PMIDs returned empty abstracts or were invalid)
- Final coverage for hypotheses 41-60: 411/422 PMIDs now have abstracts
- Result: enrichment complete — next 20 hypotheses now cite real PubMed abstracts
- Committed and pushed
2026-04-02 — Slot 0
- Fetched PubMed abstracts via NCBI eutils efetch API for top 40 hypotheses by composite score
- Collected 465+ unique PMIDs, fetched in batches of 200
- Retrieved 436 abstracts total (some PMIDs are AI-hallucinated and don't exist in PubMed)
- Updated 37 hypotheses with new abstract data
- Result: 53/149 hypotheses now have abstracts (up from 21), covering 35% of the hypothesis corpus
- Remaining 96 hypotheses with PMIDs-but-no-abstracts likely have mostly synthetic PMIDs