Literature Intake Belt
Turns HER2T’s own open questions into guarded literature mining — audit first, mine second, resolve never.
Before anything is pulled, an auditor decides whether the published literature can actually bear on the question. A “wrong instrument” verdict is recorded, not hidden — nothing gets mined.
Every claim keeps its source sentence and is provenance-tagged inferred. Mined papers enter the work queue as open — they are never treated as settled truth.
When two papers conflict, both claims are kept side by side. The belt never averages away a discordant or negative finding.
This belt only ever adds papers, claims and questions. It is mechanically incapable of closing a question — only real computed or measured evidence can do that, elsewhere.
The demand list
See the full work queueOpen questions HER2T’s own curiosity engine already flagged as possibly answerable from published papers. Running one audits it first, then mines only if the literature can genuinely bear on it.
Mine a question of your own
Ask a plain-English HER2 question. It goes through the exact same audit gate — if the literature can’t bear on it, the belt will say so and mine nothing.
Submit a specific paper to be worked
Point the belt at one paper by its PubMed ID (PMID). It goes through the exact same audit and HER2/ERBB2 relevance gates as every other mine — this is not a shortcut. If it clears them, its claims are extracted and checked for collisions. Your submission is tagged as researcher-requested for provenance.