from __future__ import annotations
from langcheck.metrics.en.reference_based_text_quality import (
semantic_similarity,
)
from langcheck.metrics.eval_clients import EvalClient
from langcheck.metrics.metric_inputs import (
get_metric_inputs,
get_metric_inputs_with_required_lists,
)
from langcheck.metrics.metric_value import MetricValue
from langcheck.metrics.scorer.detoxify_models import DetoxifyScorer
from langcheck.metrics.scorer.hf_models import (
AutoModelForSequenceClassificationScorer,
)
from langcheck.stats import compute_stats
from langcheck.utils.progress_bar import tqdm_wrapper
LANG = "en"
[docs]
def sentiment(
generated_outputs: list[str] | str,
prompts: list[str] | str | None = None,
eval_model: str | EvalClient = "local",
local_overflow_strategy: str = "truncate",
) -> MetricValue[float | None]:
"""Calculates the sentiment scores of generated outputs. This metric takes
on float values between [0, 1], where 0 is negative sentiment and 1 is
positive sentiment. (NOTE: when using an EvalClient, the sentiment scores
are either 0.0 (negative), 0.5 (neutral), or 1.0 (positive). The score may
also be `None` if it could not be computed.)
We currently support two evaluation model types:
1. The 'local' type, where the Twitter-roBERTa-base model is downloaded
from HuggingFace and run locally. This is the default model type and
there is no setup needed to run this.
2. The EvalClient type, where you can use an EvalClient typically
implemented with an LLM. The implementation details are explained in each of
the concrete EvalClient classes.
Args:
generated_outputs: The model generated output(s) to evaluate
prompts: The prompts used to generate the output(s). Prompts are
optional metadata and not used to calculate the metric.
eval_model: The type of model to use ('local' or the EvalClient instance
used for the evaluation). default 'local'
local_overflow_strategy: The strategy to handle the inputs that are too
long for the local model. The supported strategies are 'nullify',
'truncate', and 'raise'. If 'nullify', the outputs that are too long
will be assigned a score of None. If 'truncate', the outputs that
are too long will be truncated. If 'raise', an error will be raised
when the outputs are too long. The default value is 'nullify'.
Returns:
An :class:`~langcheck.metrics.metric_value.MetricValue` object
"""
metric_inputs, [generated_outputs] = get_metric_inputs_with_required_lists(
generated_outputs=generated_outputs,
prompts=prompts,
required_params=["generated_outputs"],
)
metric_name = "sentiment"
if eval_model == "local":
scores = _sentiment_local(generated_outputs, local_overflow_strategy)
explanations = None
return MetricValue(
metric_name="sentiment",
metric_inputs=metric_inputs,
explanations=explanations,
metric_values=scores,
language=LANG,
)
else: # EvalClient
assert isinstance(eval_model, EvalClient), (
"An EvalClient must be provided for non-local model types."
)
sentiment_template = eval_model.load_prompt_template(
language=LANG, metric_name=metric_name
)
sentiment_assessment_to_score = {
"Positive": 1.0,
"Neutral": 0.5,
"Negative": 0.0,
}
return eval_model.compute_metric_values_from_template(
metric_inputs=metric_inputs,
template=sentiment_template,
metric_name=metric_name,
language=LANG,
score_map=sentiment_assessment_to_score,
)
def _sentiment_local(
generated_outputs: list[str], overflow_strategy: str
) -> list[float | None]:
"""Calculates the sentiment scores of generated outputs using the
Twitter-roBERTa-base model. This metric takes on float values between
[0, 1], where 0 is negative sentiment and 1 is positive sentiment.
Ref:
https://huggingface.co/cardiffnlp/twitter-roberta-base-sentiment-latest
Args:
generated_outputs: A list of model generated outputs to evaluate
overflow_strategy: The strategy to handle inputs that are longer than
the maximum input length of the model.
Returns:
A list of scores
"""
scorer = AutoModelForSequenceClassificationScorer(
language="en",
metric="sentiment",
# Each class represents a sentiment: 0 for negative, 1 for neutral, and
# 2 for positive
class_weights=[0, 0.5, 1],
overflow_strategy=overflow_strategy,
max_input_length=512,
)
return scorer.score(generated_outputs)
[docs]
def fluency(
generated_outputs: list[str] | str,
prompts: list[str] | str | None = None,
eval_model: str | EvalClient = "local",
local_overflow_strategy: str = "truncate",
) -> MetricValue[float | None]:
"""Calculates the fluency scores of generated outputs. This metric takes on
float values between [0, 1], where 0 is low fluency and 1 is high fluency.
(NOTE: when using an EvalClient, the fluency scores are either 0.0
(poor), 0.5 (fair), or 1.0 (good). The score may also be `None` if it could
not be computed.)
We currently support two evaluation model types:
1. The 'local' type, where the Parrot fluency model is downloaded from
HuggingFace and run locally. This is the default model type and there is no
setup needed to run this.
2. The EvalClient type, where you can use an EvalClient typically
implemented with an LLM. The implementation details are explained in each of
the concrete EvalClient classes.
Args:
generated_outputs: The model generated output(s) to evaluate
prompts: The prompts used to generate the output(s). Prompts are
optional metadata and not used to calculate the metric.
eval_model: The type of model to use ('local' or the EvalClient instance
used for the evaluation). default 'local'
local_overflow_strategy: The strategy to handle the inputs that are too
long for the local model. The supported strategies are 'nullify',
'truncate', and 'raise'. If 'nullify', the outputs that are too long
will be assigned a score of None. If 'truncate', the outputs that
are too long will be truncated. If 'raise', an error will be raised
when the outputs are too long. The default value is 'nullify'.
Returns:
An :class:`~langcheck.metrics.metric_value.MetricValue` object
"""
metric_inputs, [generated_outputs] = get_metric_inputs_with_required_lists(
generated_outputs=generated_outputs,
prompts=prompts,
required_params=["generated_outputs"],
)
metric_name = "fluency"
if eval_model == "local":
scores = _fluency_local(generated_outputs, local_overflow_strategy)
explanations = None
return MetricValue(
metric_name=metric_name,
metric_inputs=metric_inputs,
explanations=explanations,
metric_values=scores,
language=LANG,
)
else: # EvalClient
assert isinstance(eval_model, EvalClient), (
"An EvalClient must be provided for non-local model types."
)
fluency_template = eval_model.load_prompt_template(
language=LANG, metric_name=metric_name
)
fluency_assessment_to_score = {
"Poor": 0,
"Fair": 0.5,
"Good": 1.0,
}
return eval_model.compute_metric_values_from_template(
metric_inputs=metric_inputs,
template=fluency_template,
metric_name=metric_name,
language=LANG,
score_map=fluency_assessment_to_score,
)
def _fluency_local(
generated_outputs: list[str], overflow_strategy: str
) -> list[float | None]:
"""Calculates the fluency scores of generated outputs using the Parrot
fluency model. This metric takes on float values between [0, 1], where 0 is
low fluency and 1 is high fluency.
Ref:
https://huggingface.co/prithivida/parrot_fluency_model
Args:
generated_outputs: A list of model generated outputs to evaluate
overflow_strategy: The strategy to handle inputs that are longer than
the maximum input length of the model.
Returns:
A list of scores
"""
scorer = AutoModelForSequenceClassificationScorer(
language="en",
metric="fluency",
# The class 1 is for fluent texts.
class_weights=[0, 1],
overflow_strategy=overflow_strategy,
)
return scorer.score(generated_outputs)
[docs]
def toxicity(
generated_outputs: list[str] | str,
prompts: list[str] | str | None = None,
eval_model: str | EvalClient = "local",
local_overflow_strategy: str = "truncate",
eval_prompt_version: str = "v2",
) -> MetricValue[float | None]:
"""Calculates the toxicity scores of generated outputs. This metric takes on
float values between [0, 1], where 0 is low toxicity and 1 is high toxicity.
(NOTE: when using an EvalClient, the toxicity scores are either 0.0
(nontoxic), or 1.0 (toxic). The score may also be `None` if it could not be
computed.)
We currently support two evaluation model types:
1. The 'local' type, where the Detoxify model is downloaded from HuggingFace
and run locally. This is the default model type and there is no setup needed
to run this.
2. The EvalClient type, where you can use an EvalClient typically
implemented with an LLM. The implementation details are explained in each of
the concrete EvalClient classes.
Args:
generated_outputs: The model generated output(s) to evaluate
prompts: The prompts used to generate the output(s). Prompts are
optional metadata and not used to calculate the metric.
eval_model: The type of model to use ('local' or the EvalClient instance
used for the evaluation). default 'local'
local_overflow_strategy: The strategy to handle the inputs that are too
long for the local model. The supported strategies are 'nullify',
'truncate', and 'raise'. If 'nullify', the outputs that are too long
will be assigned a score of None. If 'truncate', the outputs that
are too long will be truncated. If 'raise', an error will be raised
when the outputs are too long. The default value is 'nullify'.
eval_prompt_version: The version of the eval prompt to use when the
EvalClient is used. The default version is 'v2' (latest).
Returns:
An :class:`~langcheck.metrics.metric_value.MetricValue` object
"""
metric_inputs, [generated_outputs] = get_metric_inputs_with_required_lists(
generated_outputs=generated_outputs,
prompts=prompts,
required_params=["generated_outputs"],
)
metric_name = "toxicity"
if eval_model == "local":
scores = _toxicity_local(generated_outputs, local_overflow_strategy)
explanations = None
return MetricValue(
metric_name=metric_name,
metric_inputs=metric_inputs,
explanations=explanations,
metric_values=scores,
language=LANG,
)
else: # EvalClient
assert isinstance(eval_model, EvalClient), (
"An EvalClient must be provided for non-local model types."
)
toxicity_assessment_to_score = {
# The v1 prompt returns the toxicity on a scale of 1 to 5
"v1": {
"1": 0,
"2": 0.25,
"3": 0.5,
"4": 0.75,
"5": 1.0,
},
# The v2 prompt returns either "Toxic" or "Nontoxic"
"v2": {
"Toxic": 1.0,
"Nontoxic": 0,
},
}
assert eval_prompt_version in toxicity_assessment_to_score, (
f"Invalid eval_prompt_version: {eval_prompt_version}. The valid versions are {list(toxicity_assessment_to_score.keys())}."
)
toxicity_template = eval_model.load_prompt_template(
language=LANG,
metric_name=metric_name,
eval_prompt_version=eval_prompt_version,
)
return eval_model.compute_metric_values_from_template(
metric_inputs=metric_inputs,
template=toxicity_template,
metric_name=metric_name,
language=LANG,
score_map=toxicity_assessment_to_score[eval_prompt_version],
)
def _toxicity_local(
generated_outputs: list[str], overflow_strategy: str
) -> list[float | None]:
"""Calculates the toxicity scores of generated outputs using the Detoxify
model. This metric takes on float values between [0, 1], where 0 is low
toxicity and 1 is high toxicity.
Ref:
https://github.com/unitaryai/detoxify
Args:
generated_outputs: A list of model generated outputs to evaluate
Returns:
A list of scores
"""
return DetoxifyScorer(overflow_strategy=overflow_strategy).score(
generated_outputs
)
[docs]
def flesch_reading_ease(
generated_outputs: list[str] | str,
prompts: list[str] | str | None = None,
) -> MetricValue[float]:
"""Calculates the readability of generated outputs using the Flesch Reading
Ease Score. This metric takes on float values between (-∞, 121.22], but
typically ranges between 0 and 100, where higher scores mean the text is
easier to read.
The score is based on the number of sentences, words, and syllables in the
text. See "How to Write Plain English" by Rudolf Franz Flesch for more
details.
Args:
generated_outputs: The model generated output(s) to evaluate
prompts: The prompts used to generate the output(s). Prompts are
optional metadata and not used to calculate the metric.
Returns:
An :class:`~langcheck.metrics.metric_value.MetricValue` object
"""
metric_inputs, [generated_outputs] = get_metric_inputs_with_required_lists(
generated_outputs=generated_outputs,
prompts=prompts,
required_params=["generated_outputs"],
)
output_stats = [
compute_stats(output)
for output in tqdm_wrapper(generated_outputs, desc="Computing stats")
]
scores = [
206.835
- 1.015 * (stat.num_words / stat.num_sentences)
- 84.6 * (stat.num_syllables / stat.num_words)
for stat in output_stats
]
return MetricValue(
metric_name="flesch_reading_ease",
metric_inputs=metric_inputs,
explanations=None,
metric_values=scores,
language="en",
)
[docs]
def flesch_kincaid_grade(
generated_outputs: list[str] | str,
prompts: list[str] | str | None = None,
) -> MetricValue[float]:
"""Calculates the readability of generated outputs using the Flesch-Kincaid
Grade Level metric. This metric takes on float values between [-3.40, ∞),
but typically ranges between 0 and 12 (corresponding to U.S. grade levels),
where lower scores mean the text is easier to read.
Like the Flesch Reading Ease Score, this metric is based on the number of
sentences, words, and syllables in the text.
Ref:
https://apps.dtic.mil/sti/citations/ADA006655
Args:
generated_outputs: The model generated output(s) to evaluate
prompts: The prompts used to generate the output(s). Prompts are
optional metadata and not used to calculate the metric.
Returns:
An :class:`~langcheck.metrics.metric_value.MetricValue` object
"""
metric_inputs, [generated_outputs] = get_metric_inputs_with_required_lists(
generated_outputs=generated_outputs,
prompts=prompts,
required_params=["generated_outputs"],
)
output_stats = [
compute_stats(output)
for output in tqdm_wrapper(generated_outputs, desc="Computing stats")
]
scores = [
0.39 * (stat.num_words / stat.num_sentences)
+ 11.8 * (stat.num_syllables / stat.num_words)
- 15.59
for stat in output_stats
]
return MetricValue(
metric_name="flesch_kincaid_grade",
metric_inputs=metric_inputs,
explanations=None,
metric_values=scores,
language="en",
)
[docs]
def ai_disclaimer_similarity(
generated_outputs: list[str] | str,
prompts: list[str] | str | None = None,
ai_disclaimer_phrase: str = (
"I don't have personal opinions, emotions, or consciousness."
),
eval_model: str | EvalClient = "local",
) -> MetricValue[float]:
"""Calculates the degree to which the LLM's output contains a disclaimer
that it is an AI. This is calculated by computing the semantic similarity
between the generated outputs and a reference AI disclaimer phrase; by
default, this phrase is "I don't have personal opinions, emotions, or
consciousness.", but you can also pass in a custom phrase. Please refer to
:func:`~langcheck.eval.en.reference_based_text_quality.semantic_similarity`
for details on the typical output ranges and the supported embedding model
types.
Args:
generated_outputs: A list of model generated outputs to evaluate
prompts: An optional list of prompts used to generate the outputs.
Prompts are not evaluated and only used as metadata.
ai_disclaimer_phrase: Reference AI disclaimer phrase, default "I don't
have personal opinions, emotions, or consciousness."
eval_model: The type of model to use ('local' or the EvalClient instance
used for the evaluation). default 'local'
Returns:
An :class:`~langcheck.metrics.metric_value.MetricValue` object
"""
metric_inputs, [generated_outputs] = get_metric_inputs_with_required_lists(
generated_outputs=generated_outputs,
prompts=prompts,
required_params=["generated_outputs"],
)
ai_disclaimer_phrase_list = [ai_disclaimer_phrase] * len(generated_outputs)
semantic_similarity_values = semantic_similarity(
generated_outputs, ai_disclaimer_phrase_list, prompts, eval_model
)
return MetricValue(
metric_name="ai_disclaimer_similarity",
metric_inputs=metric_inputs,
explanations=None,
metric_values=semantic_similarity_values.metric_values,
language="en",
)
[docs]
def jailbreak_prompt(
prompts: list[str] | str,
eval_model: EvalClient,
) -> MetricValue[float | None]:
"""Calculates whether jailbreak techniques are included in the prompts.
This metric takes on float values of either 0.0 (Low Risk),
0.5 (Medium Risk), or 1.0 (High Risk). The score may also be `None`
if it could not be computed.
We currently only support the evaluation based on an EvalClient.
"""
metric_inputs = get_metric_inputs(
prompts=prompts,
required_params=["prompts"],
)
metric_name = "jailbreak_prompt"
jailbreak_prompt_template = eval_model.load_prompt_template(
language=LANG, metric_name=metric_name
)
return eval_model.compute_metric_values_from_template(
metric_inputs=metric_inputs,
template=jailbreak_prompt_template,
metric_name=metric_name,
language=LANG,
score_map={
"Low Risk": 0.0,
"Medium Risk": 0.5,
"High Risk": 1.0,
},
)
[docs]
def prompt_leakage(
generated_outputs: list[str] | str,
system_prompts: list[str] | str,
eval_model: EvalClient,
eval_prompt_version: str = "v2",
) -> MetricValue[float | None]:
"""Calculates the severity of prompt leakage in the generated outputs.
This metric takes on float values of either 0.0 (Low Risk),
0.5 (Medium Risk), or 1.0 (High Risk). The score may also be `None`
if it could not be computed.
We currently only support the evaluation based on an EvalClient.
"""
if eval_prompt_version not in ["v1", "v2"]:
raise ValueError(
f"Invalid eval_prompt_version: {eval_prompt_version}. The valid versions are 'v1' and 'v2'."
)
metric_inputs = get_metric_inputs(
generated_outputs=generated_outputs,
additional_inputs={
"system_prompts": system_prompts,
},
additional_input_name_to_prompt_var_mapping={
"system_prompts": "system_prompt",
},
required_params=["generated_outputs", "system_prompts"],
)
metric_name = "prompt_leakage"
prompt_leakage_template = eval_model.load_prompt_template(
language=LANG,
metric_name=metric_name,
eval_prompt_version=eval_prompt_version,
)
return eval_model.compute_metric_values_from_template(
metric_inputs=metric_inputs,
template=prompt_leakage_template,
metric_name=metric_name,
language=LANG,
score_map={
"Low Risk": 0.0,
"Medium Risk": 0.5,
"High Risk": 1.0,
},
)
