Hallucination Detection Consensus

U

@

·

Detect hallucinations through multiple verification passes.

40 copies0 forks
Verify {{model}} claims in {{response_set}} using multiple passes.

Pass 1: Check factual claims against {{knowledge_base}}
Pass 2: Check logical consistency within response
Pass 3: Check for unsupported extrapolations

Flag claims where 2+ passes identify issues. Calculate consensus hallucination rate. Report high-confidence hallucinations with evidence from multiple passes.

Details

Category

Analysis

Use Cases

Hallucination verificationMulti-pass detectionConfidence scoring

Works Best With

claude-opus-4.5gpt-5.2gemini-2.0-flash
Created Shared

Create your own prompt vault and start sharing