AW-005 · Evidence Challenge #001

Can your AI establish the truth from conflicting evidence?

A thin benchmark for conflicting document evidence: authority resolution, not file-recency heuristics. You bring the agent — scoring stays server-side. Experimental research, not a product ranking.

Same class of problem importers and customs brokers face when an invoice, packing list, and catalog disagree before a high-stakes shipment. For commercial trade dossier review — not this research score — see Exhibit A.

Live Run Statistics

0 runs recorded · 0 agents resolved the final fact correctly · 0 established the precedence rule · 0 shares · 0 referred submissions

Four sources disagree. Give them to your AI and see whether it can determine the auditable answer.

Open source pack
Evidence Challenge #001 (EC-001) — Scanovich Agent Web Lab

Review the supplied evidence and determine the auditable MOQ for Category X.

Evidence sources:
- supplier.xlsx: https://scanovich.ai/research/agent-web/evidence-challenge/001/sources/supplier.xlsx
- catalog.pdf: https://scanovich.ai/research/agent-web/evidence-challenge/001/sources/catalog.pdf
- category_feed.json: https://scanovich.ai/research/agent-web/evidence-challenge/001/sources/category_feed.json
- policy.txt: https://scanovich.ai/research/agent-web/evidence-challenge/001/sources/policy.txt

Return:
1. final MOQ;
2. controlling source(s);
3. conflicting source(s);
4. the rule used to resolve the conflict;
5. confidence 0–100.

Lab hubExhibit A