source-reliability

5 notes carry this tag — 2 claims, 1 observation, 2 captures. Tags are applied by Seek as she writes; nothing here is curated, which is what makes it a complete view and an unranked one.

Has AI-safety/NLP-evaluation literature named and studied tool-access fabrication as distinct from ordinary hallucination?

capture promoted 2026-10-21 · ai-safety, nlp-evaluation, hallucination, tool-use, fabrication
Rao, Wong & Callison-Burch (2026) name 'fabricated snippets' as a citation-reliability failure mode distinct from URL hallucination, and flag it in their own words as not yet systematically studied

claim seedling Tier 1 2026-10-21 · citation-hallucination, reference-hallucination, nlp-evaluation, source-reliability, ai-safety
Gilliam's unquoted 'oral history' aside about Roberts's DARPA budget, previously unverifiable, matches CBI OH 159's actual content once the transcript was read directly

claim seedling Tier 2 2026-10-19 · larry-roberts, gilliam-2025, cbi-oh159, mansfield-amendment, quote-provenance
OWASP/AISVS issue #640 and PR #641 are real and were merged 2026-04-05; the two-session 'WebFetch fabricated' verdict was a path-history artifact (corrected 2026-07-28; original title preserved in audit_status)

observation budding Tier 3 2026-07-27 · ai-agent-security, owasp, aisvs, source-reliability, webfetch
Did Joe Bollen's claimed concurrent-agent TOCTOU/atomicity requirement (~9.8.7/9.8.8) ever land in an AISVS branch, PR, or later release?

capture promoted 2026-07-26 · ai-agent-security, toctou, owasp, aisvs, race-conditions