Implementation:Explodinggradients Ragas RubricsScore Metric

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

Overview

RubricsScore evaluates submissions against domain-specific rubrics on a 1-5 discrete scale, supporting both single-turn and multi-turn interactions.

Description

The RubricsScore class uses an LLM to score submissions based on configurable rubric definitions. Two default rubric sets are provided: DEFAULT_REFERENCE_FREE_RUBRICS (evaluates against user input only) and DEFAULT_WITH_REFERENCE_RUBRICS (evaluates against a reference answer). The rubrics text is injected into the prompt instruction during construction. The class supports both single-turn and multi-turn evaluation modes. It inherits from MetricWithLLM, SingleTurnMetric, and MultiTurnMetric.

Key attributes:

• rubrics -- A dictionary mapping score descriptions (e.g., "score1_description" through "score5_description") to their criteria text.
• single_turn_scoring_prompt / multi_turn_scoring_prompt -- Customizable prompts with rubrics injected.
• max_retries -- Maximum LLM retries (default 1).

Usage

All input columns are optional. The metric requires an LLM. Rubrics can be customized at construction time.

Code Reference

Property	Value
Source Location	src/ragas/metrics/_domain_specific_rubrics.py L85-177
Class Signature	class RubricsScore(MetricWithLLM, SingleTurnMetric, MultiTurnMetric)
Import	from ragas.metrics import RubricsScore

I/O Contract

Inputs

Parameter	Type	Required	Description
user_input	str	Optional	The user query or input
response	str	Optional	The generated response
retrieved_contexts	List[str]	Optional	Retrieved context passages
reference	str	Optional	The reference answer
reference_contexts	List[str]	Optional	Reference context passages

Outputs

Output	Type	Description
score	int	Discrete score from 1 to 5 based on rubric evaluation

Usage Examples

from ragas.metrics import RubricsScore
from ragas.metrics._domain_specific_rubrics import DEFAULT_WITH_REFERENCE_RUBRICS
from ragas.dataset_schema import SingleTurnSample

metric = RubricsScore(
    name="quality_rubric",
    rubrics=DEFAULT_WITH_REFERENCE_RUBRICS
)
# metric.llm = ...  # Set your LLM

sample = SingleTurnSample(
    user_input="Explain photosynthesis.",
    response="Photosynthesis converts sunlight into chemical energy in plants.",
    reference="Photosynthesis is the process by which plants convert light energy into chemical energy, producing glucose and oxygen from carbon dioxide and water."
)
# score = await metric.single_turn_ascore(sample)

Related Pages

• Explodinggradients_Ragas_InstanceRubrics_Metric -- Per-sample rubric evaluation (rubrics provided per instance)
• Explodinggradients_Ragas_SimpleCriteriaScore_Metric -- User-defined criteria with discrete scoring
• Explodinggradients_Ragas_AspectCritic_Metric -- Binary aspect-based evaluation