Detecting Contradictions Across Retrieved Sources
Identifies conflicts among retrieved legal sources — versions, amendments, contradictory clauses — and classifies the conflict type instead of silently picking one.
Prompt
You analyse consistency across legal sources. An undetected contradiction is a serious failure: the model will silently pick one source and the user will never know.
Sources: <sources> [Paste chunks with id, document name, and date where available] </sources>
Identify every conflict and classify it:
- VERSION — the same clause across different versions of one document. The later version prevails where dates are known.
- AMENDMENT — a later provision that amended or repealed an earlier one.
- HIERARCHY — norms at different levels (statute vs regulation vs internal guidance).
- SCOPE — two rules that look contradictory but apply to different situations. Not a real contradiction — explain the distinction.
- GENUINE — a substantive conflict not resolvable from the sources themselves. Requires human determination.
For each conflict return: { "source_a": "id", "source_b": "id", "type": "", "quote_a": "exact quote", "quote_b": "exact quote", "resolution": "how to resolve, or 'requires human determination'", "impact": "what changes in the user-facing answer under each side" }
If no conflicts exist return an empty list. Do not manufacture contradictions from different phrasings of the same rule.
Why this converts well
- Ready to copy and use immediately
- Tailored for legal and public-sector work
- Easy to adapt to internal workflows
Tags
Related Prompts
Strict Source-Grounded Answering over a Legal Corpus
System prompt that forbids parametric knowledge, requires a cited source span for every claim, and defines explicit behaviour when the retrieved context does not cover the question.
Chunking Strategy Design for Legal Documents
Analyses a legal document type and produces a justified chunking strategy — boundaries, overlap, metadata, and edge cases — instead of arbitrary fixed-size splitting.
Building an Evaluation Set for a Legal Retrieval System
Generates an evaluation set from a given corpus, including questions that should fail, to measure retrieval and faithfulness rather than only happy-path answering.