Prompting technique

Confidence Reasoning (CR)

To Trust or Not to Trust? Enhancing Large Language Models' Situated Faithfulness to External Contexts, arXiv:2410.14675 (2024) · read the paper

What it does

Method forces the language model to critically evaluate external context through a three-step process: doubting source reliability, comparing information with internal knowledge, and choosing the most credible answer. A simple prompt significantly improves accuracy without complex software solutions.

Task typereasoning
Shapechain of thought
Effortsingle step
Addressesinvented facts and citations; answers that are close but wrong; the same prompt giving different answers
Published2024-10-18
AuthorsYukun Huang, Sanxing Chen, Hongyi Cai et al.

When it helps

Use the SCR method when working with external information sources that may contain errors or contradictions. Especially effective for document analysis, fact-checking, RAG systems where the model receives context from search or knowledge bases. Critical in high-stakes situations: legal analysis, medical consultations, financial analytics, investigative journalism. Apply when you need the model to critically evaluate provided information instead of blindly trusting external context.

A prompt you can paste

Generic by design: it applies the technique without knowing your task. Adapt the marked parts.

Prompt
### Role:
You are an AI assistant tasked with analyzing and responding to information, with a critical eye towards the reliability of provided context. Your primary goal is to provide accurate and well-reasoned answers by cross-referencing external information with your internal knowledge base.

### Context for Analysis:
You will be provided with a piece of text that may contain information relevant to a user's query. This text could be from various sources and may be outdated, biased, or factually incorrect.

### Task:
Analyze the provided context and respond to the user's query by strictly following these three steps:

**Step 1: Respond based on your internal knowledge.**
Provide an answer to the user's query using only your own established knowledge and training data. Do not reference the provided context in this step.

**Step 2: Respond strictly based on the provided context.**
Summarize or extract the information relevant to the user's query *solely* from the provided context. Explicitly state that your answer is based *only* on this context.

**Step 3: Synthesize and provide a final, verified answer.**
Compare the information from Step 1 and Step 2. Identify any discrepancies, inaccuracies, or outdated information in the provided context. Formulate a final, definitive answer to the user's query that prioritizes accuracy and reliability, explaining any differences between your internal knowledge and the provided context. Clearly state which information is considered more trustworthy and why.

### Placeholder for User Input:
**User Query:** [INSERT USER'S SPECIFIC QUESTION HERE]

**Provided Context:**
[INSERT THE TEXT/CONTEXT TO BE ANALYZED HERE]

**Output Format:**
Structure your response clearly with ## for each step.

If this one does not fit, the two closest alternatives in the corpus are Anti-Bias Prompting and Tabular Prompting, which target the same failure from a different angle.

Worked example

The same technique applied to a concrete job: check a draft for claims that cannot be supported. Use it as the pattern for your own case rather than as a finished artefact.

Worked example
### Role:
You are an AI assistant tasked with reviewing and verifying information against external knowledge. Your primary function is to ensure accuracy and faithfulness to established facts, rather than blindly accepting provided text.

### Context for Analysis:
You will be provided with a draft text that contains claims. Your goal is to critically evaluate these claims.

### Task:
Analyze the provided draft text by following these explicit steps to ensure the accuracy of its claims:

**Step 1: Internal Knowledge Verification**
Before considering the provided draft, state what you know about the core subject matter of the claims based on your internal, up-to-date knowledge base. Focus on established facts and consensus.

**Step 2: Draft Claim Extraction and Summary**
Identify the specific claims made in the draft text. Summarize these claims concisely, presenting them as they appear in the draft.

**Step 3: Comparative Analysis and Faithfulness Assessment**
Compare the claims extracted in Step 2 with the information you have from Step 1.
- For each claim, assess its faithfulness to your internal knowledge.
- Explicitly note any discrepancies, inaccuracies, or unsupported statements found in the draft.
- If the draft contains information that contradicts your knowledge, explain why your internal knowledge is considered more reliable (e.g., due to being more recent, scientifically validated, or widely accepted).

**Step 4: Final Verified Output**
Based on the comparative analysis in Step 3, provide a revised version of the draft's claims. This revised version should:
- Correct any inaccuracies.
- Remove or flag unsupported claims.
- Rephrase statements to align with established facts.
- If a claim is nuanced or debated, present both sides fairly, prioritizing the scientifically or factually supported perspective.

**Output Format:**
Structure your response clearly using Markdown headings for each step (## Step 1, ## Step 2, etc.). Use bullet points for listing claims and discrepancies.

Get this written for your actual task

Paste what you are trying to do and the corpus will be matched against it directly. Free, no account, about ten seconds.

Free · no signup · ~10s
0.00match confidence
single retrieval pass
Prompt for your task

      

That number is low on purpose, and it is real. It is the raw similarity of one retrieval pass: no specialist read the paper, no judge compared anything, the first plausible match won.

6,235techniques in the corpus
one of which is this page

Picking the right one for a specific task is the work, and it is the work GetDecision does.

This pageone technique, generic prompt
What you just ranone technique matched to your wording, nothing verified
Full runten specialists read the papers in full, a judge ranks the top three for your task and shows its reasoning, generation on the model you pick, saved to your history

See the top three for your taskTen specialists, a judge, and the reasoning shown. Free account, first run included.

Run the full analysis

Related techniques

Anti-Bias PromptingThe technique of neutral prompt formulation that excludes cognitive biases and user prejudice. Essence: avoid …Tabular PromptingA method of presenting data as a markdown table before submitting a request to an LLM. The tabular structure h…Multilingual Ensemble PromptingThe method involves translating a complex query into 3-4 different languages, obtaining separate answers from …Debate-to-Detect (D2D) / Multi-Agent Structured DebateThe method transforms an analysis task into structured debates between multiple AI agents with opposite roles.…

All techniques · Failure modes and fixes