Hasan Iqbal
Added LLM Evaluation on Datasets
eca534f unverified
raw
history blame
1.41 kB
import os
import yaml
from .factool_utils.chat_api import OpenAIChat
from .factool_utils.prompt import CLAIM_EXTRACTION_PROMPT
from openfactcheck.core.solver import StandardTaskSolver, Solver
from openfactcheck.core.state import FactCheckerState
@Solver.register("factool_decontextualizer", "response", "claims")
class FactoolDecontextualizer(StandardTaskSolver):
"""
A solver to extract claims from a response.
"""
def __init__(self, args):
super().__init__(args)
self.gpt_model = self.global_config.get("llm_in_use", "gpt-4o")
self.gpt = OpenAIChat(self.gpt_model)
self.claim_prompt = CLAIM_EXTRACTION_PROMPT
def __call__(self, state: FactCheckerState, *args, **kwargs):
response = state.get(self.input_name)
claims = self._claim_extraction(responses=[response])[0]
extracted_claims = [claim["claim"] for claim in claims]
state.set(self.output_name, extracted_claims)
return True, state
def _claim_extraction(self, responses):
messages_list = [
[
{"role": "system", "content": self.claim_prompt["system"]},
{
"role": "user",
"content": self.claim_prompt["user"].format(input=response),
},
]
for response in responses
]
return self.gpt.run(messages_list, list)