Implementation:Explodinggradients Ragas Collections ContextPrecision Metric

Field	Value
source	Repo
domains	Metrics, Evaluation
last_updated	2026-02-10 00:00 GMT

Overview

The ContextPrecision module provides four v2 class-based metrics -- ContextPrecisionWithReference, ContextPrecisionWithoutReference, ContextPrecision, and ContextUtilization -- that evaluate whether retrieved contexts are useful for answering a question using LLM-based average precision scoring.

Description

All classes extend BaseMetric and require a modern InstructorBaseRagasLLM for evaluation.

ContextPrecisionWithReference evaluates each retrieved context against a reference answer. For each context, it uses a structured prompt (ContextPrecisionPrompt) to ask the LLM whether the context is useful for producing the reference answer, yielding a binary verdict (0 or 1). The final score is computed as average precision across all verdicts, rewarding retrieval systems that rank useful contexts higher.

ContextPrecisionWithoutReference uses the same approach but compares contexts against the generated response instead of a reference answer.

ContextPrecision is a convenience alias for ContextPrecisionWithReference with the default name "context_precision".

ContextUtilization is a convenience alias for ContextPrecisionWithoutReference with the default name "context_utilization".

The average precision algorithm accumulates a cumulative sum of verdicts and computes precision at each relevant position, then divides by the total number of relevant contexts.

Usage

Instantiate with a required llm parameter. Use the WithReference variant when you have ground truth answers, and the WithoutReference variant when only the generated response is available.

Code Reference

Property	Value
Source Location	src/ragas/metrics/collections/context_precision/metric.py L1--330
Signatures	class ContextPrecisionWithReference(BaseMetric), class ContextPrecisionWithoutReference(BaseMetric), class ContextPrecision(ContextPrecisionWithReference), class ContextUtilization(ContextPrecisionWithoutReference)
Import	from ragas.metrics.collections import ContextPrecision

I/O Contract

Inputs (ContextPrecisionWithReference / ContextPrecision)

Parameter	Type	Required	Description
user_input	str	Yes	The question being asked
reference	str	Yes	The reference / ground truth answer
retrieved_contexts	List[str]	Yes	Retrieved contexts to evaluate

Inputs (ContextPrecisionWithoutReference / ContextUtilization)

Parameter	Type	Required	Description
user_input	str	Yes	The question being asked
response	str	Yes	The generated response
retrieved_contexts	List[str]	Yes	Retrieved contexts to evaluate

Constructor Parameters

Parameter	Type	Default	Description
llm	InstructorBaseRagasLLM	(required)	Modern instructor-based LLM for context evaluation
name	str	varies by class	Metric name

Outputs

Field	Type	Description
MetricResult.value	float	Average precision score in range 0.0--1.0 (higher is better)

Usage Examples

import openai
from ragas.llms.base import llm_factory
from ragas.metrics.collections import ContextPrecision

# Setup
client = openai.AsyncOpenAI()
llm = llm_factory("gpt-4o-mini", client=client)

# ContextPrecision (with reference)
metric = ContextPrecision(llm=llm)
result = await metric.ascore(
    user_input="What is the capital of France?",
    reference="Paris is the capital of France.",
    retrieved_contexts=[
        "Paris is the capital and largest city of France.",
        "Berlin is the capital of Germany.",
        "France is a country in Western Europe.",
    ]
)
print(f"Context Precision: {result.value}")

# ContextUtilization (without reference)
from ragas.metrics.collections.context_precision.metric import ContextUtilization

util_metric = ContextUtilization(llm=llm)
result = await util_metric.ascore(
    user_input="What is the capital of France?",
    response="Paris is the capital of France.",
    retrieved_contexts=["Paris is the capital of France.", "Unrelated context."]
)
print(f"Context Utilization: {result.value}")

Related Pages

• Explodinggradients_Ragas_Collections_BaseMetric_Class -- Base class for all v2 metrics
• Explodinggradients_Ragas_Collections_ContextRelevance_Metric -- Context relevance with dual-judge evaluation
• Explodinggradients_Ragas_Collections_AnswerRelevancy_Metric -- Answer relevancy evaluation
• Explodinggradients_Ragas_Collections_ResponseGroundedness_Metric -- Response groundedness evaluation