VERIFIED EVIDENCE LEDGER (human-readable program extraction, provided now; these facts were missing from your prior compact CSV):
F1 Samsung source samsung-2025-consolidated.txt contains capture header 2026-10-14 and official DART receipt 20260310002820. Header describes acquisition date, not guarantee of no subsequent corrections. Actual report FY2025 has 2023-25 comparative statements.
F2 Samsung 2025 total net income 45,206,805; owners 44,260,956; noncontrolling945,849. Exact source: consolidated income statement and net-income attribution.
F3 LG 2025 SG&A18,363,266 vs2024 17,959,559; gross profit20,841,658 vs21,379,234. Verified in lg-2025-consolidated.txt consolidated income statement. Thus operating-profit bridge: gross-profit delta -537,576 and SG&A delta +403,707 -> op-profit delta -941,283. This is an accounting bridge, NOT proof tariffs caused the changes.
F4 Sources actually given to Gemini were full TXT including this evidence; Grok and your first run were given compact 54-number CSV plus earlier outputs. 'Not in reviewer input' is NOT 'not in original document'.
F5 The statement 'semiconductor recovery may have caused growth' is not proved by the supplied statements and was unsolicited. Remove from approved findings, retain as unverified question only.
F6 LG CFO4,280,500 vs3,842,661; net income1,220,412 vs591,365; op2,478,392 vs3,419,675.
F7 Jev 4 selected source cases in original and reversed order yielded same correct answer each (8 requests), narrow curated educational set, not general accuracy evaluation. Confidence profit .51/.55 cash .55/.43 loss .99/.97 cause NONE .93/.93. No universal threshold claim.
F8 No one checked later amendments or current 2026 financial performance; none of these runs had browsing tools.