What does the Grounding Plugin — Reference Check Entities do?

It checks whether entities in a given list are actually supported by a supplied reference context, and returns a bulleted list of any entities that are NOT backed by that reference material.

When should I use this prompt?

Use it as a verification step after extracting or generating entities to catch hallucinations or unsupported claims in RAG or fact-checking pipelines built on Semantic Kernel.

What are the inputs and outputs?

Inputs are the list of entities to check and the reference context to check them against; the output is a bulleted list of entities not supported by the reference context.

Can this be used as a standalone grounding solution?

No, it's one plugin within Semantic Kernel's GroundingPlugin sample set and is meant to be composed with entity-extraction and other grounding-check prompts rather than used in isolation.

Prompt

Verify Claims Against Reference Context

Name: Grounding Plugin - Reference Check Entities
Availability: OnlineOnly
Author: Semantic Kernel

Semantic Kernel prompt template that flags which entities in a list aren't supported by a given reference context.

Copy prompt

Semantic Kernel

Maintainer?

Spark score

out of 100

Updated last month

Version python-1.43.1

Add to Favorites

Why it matters

Ensure all claims made within a given text are factually supported by a provided reference document. This asset identifies and lists any unsupported statements, helping to maintain accuracy and credibility.

Outcomes

What it gets done

Check if entities are grounded in reference context.

Return a bulleted list of unsupported entities.

Verify factual accuracy of claims.

Install

Add it to your toolbox

Run in your project directory:

curl -fsSL https://spark.entire.vc/get/sk-groundingplugin-referencecheckentities | bash

Overview

Grounding Plugin - Reference Check Entities

A Semantic Kernel prompt template that checks a list of entities against a reference context and flags unsupported ones. Use as a verification step in a grounding or fact-checking pipeline, composed with other GroundingPlugin prompts, not standalone.

What it does

This Semantic Kernel prompt template checks whether a given list of entities is actually grounded in a supplied reference context. Any entities not supported by that reference context are returned as a bulleted list, making it a targeted grounding-check step within a larger grounding pipeline.

When to use - and when NOT to

Use this as a verification step after extracting or generating a list of entities, to catch cases where an entity was hallucinated or isn't actually backed by the source material you're grounding against - a common need in RAG or fact-checking pipelines built on Semantic Kernel.

Not a full grounding pipeline on its own - it's one plugin in Semantic Kernel's broader GroundingPlugin sample set, meant to be composed with entity-extraction and other grounding-check prompts rather than used in isolation.

Inputs and outputs

Inputs: input (the list of entities to check) and reference_context (the source text to check them against).

Outputs: a bulleted list of entities from input that are NOT supported by reference_context.

Integrations

Microsoft Semantic Kernel, as part of its GroundingPlugin prompt template samples.

Who it's for

Developers building Semantic Kernel-based grounding or fact-checking pipelines who need a discrete step for flagging unsupported entities.

Source README

Check to see if a given list of entities is grounded in a reference context. Any of the items which are not supported by the reference context will be returned as a bulleted list.

Input variables: input, reference_context

FAQ

Common questions

Discussion