Model Card for Model ID
This is the MCQ confidence prediction model that outputs a certainty score given a math question and a full step-by-step rationale that attempts to solve the question.
Model Details
Model Description
- Developed by: Shiyao Li
- Finetuned from model : meta-llama/Meta-Llama-3.1-8B
Model Example Usage
from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch
model_name_or_path = 'CarelessLee/MCQ_pooled_full_rationale_confidence_predictor'
model = AutoModelForSequenceClassification.from_pretrained(model_name_or_path)
tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)
def predict(text):
model.eval()
inputs = tokenizer(text, return_tensors="pt")
with torch.no_grad():
outputs = model(**inputs)
logits = outputs.logits
certainty_score = logits.item()
return certainty_score
def evaluate_predictor(label_data):
for sample in tqdm(label_data, desc="Processing questions"):
highest_score = -1
best_rationale = ""
for rationale in sample['rationales']:
text = f"Problem: {sample['question']}\n---\nRationale Step: {rationale}"
predicted_certainty_score = predict(text)
print("predicted_certainty_score: ", predicted_certainty_score)
if __name__ == "__main__":
with open("example.json", 'r') as f:
label_data = json.load(f)
evaluate_predictor(label_data)
- Downloads last month
- 4,306
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.