Loading patterns…
Chain of Verification (CoVe)
Reduces hallucination by drafting a baseline answer, planning a set of verification questions about it, answering each of those questions independently so the checks are not biased by the draft, and then revising the final answer using only the verified facts. Factored verification, where checks run in separate contexts, outperforms letting the model re-check its own draft, which tends to repeat its own mistakes.
In 30 seconds
- What
- Drafts an answer, generates verification questions about it, answers each question independently without seeing the draft, then revises using only verified facts.
- When to use
- Factual tasks prone to hallucination where you can decompose claims into checkable questions and tolerate moderate latency overhead.
- Watch out
- Verification questions inherit the draft's errors if they're poorly designed; independent answering adds 2-3x cost and latency.
Ask the AI expert about this pattern
Opens the assistant with your question prefilled. You review it before sending.
Chain of Verification (CoVe): Overview
Reduces hallucination by drafting a baseline answer, planning a set of verification questions about it, answering each of those questions independently so the checks are not biased by the draft, and then revising the final answer using only the verified facts. Factored verification, where checks run in separate contexts, outperforms letting the model re-check its own draft, which tends to repeat its own mistakes.
- Generates a baseline draft answer first
- Plans targeted verification questions about the draft
- Answers each verification question independently of the draft
- Factored execution stops the draft from biasing its own checks
- Revises the final answer to drop unverified claims
- Cuts hallucination on list-based and closed-book factual tasks
The Agent Architect
One pattern, one tradeoff, one production failure story. A short weekly briefing for people building agentic systems.
Weekly email, one-click unsubscribe. We only use your address to send the briefing.
References
The papers, specifications, and repositories this pattern is based on.
- Chain-of-Verification Reduces Hallucination in Large Language Models - Dhuliawala et al. (ArXiv 2023)arXiv:2309.11495
- Chain-of-Verification (CoVe), Meta AI - Findings of ACL 2024
From the engineer behind this catalog
Get your agent architecture reviewed
This page documents one pattern. Your system runs dozens, and most failures live in how they fit together. Have the whole design reviewed against the 288 patterns in this catalog: architecture, reliability, evaluation and cost, every finding mapped to the pattern that fixes it.
€750 instead of €1,500, one week, written report and walkthrough call, until 30 September