F-2395Detected Oct 6, 16:52 UTC
How do I turn on two-factor authentication for My Harbor?
Found in v3.4 · Gemini 2.5 Pro comparison on support-golden-v4 · Account & security
Model answer vs gold answer
54% of model-answer words appear in the gold answer
Model answer
v3.4 · Gemini 2.5 ProGo to My Harbor > Security > Two-factor authentication and choose an authenticator app or SMS. Save your backup codes somewhere safe. Note: Unsupported: Harbor support can disable 2FA for you over chat if you lose your phone.
Gold answer
Human-verified · support-golden-v4Go to My Harbor > Security > Two-factor authentication and choose an authenticator app or SMS. Save your backup codes. If you lose your device, use a backup code or complete identity verification through account recovery.
- claimFlagged by the judge as unsupported or contradicted
Retrieved context
bge-m3 · hybrid (BM25 + dense, RRF) · top 3 passed to the generator
3 of 3 retrieved chunks are relevant · gold passage ranked first
- 1
Two-factor authentication for My Harbor
Turning on 2FA · kb-account-011 · v2026.09
RelevantOpen My Harbor > Security > Two-factor authentication.
Similarity0.92 - 2
Two-factor authentication for My Harbor
Backup codes · kb-account-011 · v2026.09
RelevantSave your backup codes when you turn on 2FA.
Similarity0.87 - 3
Two-factor authentication for My Harbor
Lost device · kb-account-011 · v2026.09
RelevantWithout backup codes, complete account recovery. 2FA cannot be removed over chat.
Similarity0.83
Judge verdict
LLM judge · Claude Opus 5.5 · rubric v3
0.10
rubric score · pass threshold 0.70
The final sentence contradicts chunk 3 and describes a social-engineering path the policy forbids.
Diagnosis
Unsupported claims · SIM swap and fraud
Root cause
Generator hallucinated a support shortcut on a security intent.
Suggested fix
Block 'disable 2FA' phrasing in output filters; add security intents to the strict-grounding route.
Playbook: unsupported claims
The answer states something that is not present in the retrieved context.
Detection signals
- Claim-level faithfulness below 0.85
- Numbers, fees or durations absent from every retrieved chunk
- Citations that do not contain the cited sentence
Common fixes
- Require a citation after each factual sentence and refuse when none applies
- Reject answers with numeric claims missing from context in a post-generation check
- Lower temperature for billing and policy intents
- Add the query to the regression set with an escalate-or-decline gold answer
Activity
1 event · 0 comments
Relevant ·
Detected by v3.4 · Gemini 2.5 Pro run and labeled unsupported claims