metadata

license: apache-2.0

Cross-Encoder for Hallucination Detection

This model was trained using SentenceTransformers Cross-Encoder class. This model is based on microsoft/deberta-v3-base.

Training Data

The model was trained on the NLI data and a variety of datasets evaluating summarization accuracy for factual consistency, including FEVER, Vitamin C and PAWS.

Performance

TRUE Dataset (Minus Vitamin C, FEVER and PAWS) - 0.872 AUC Score
SummaC Benchmark (Test Split) - 0.764 Balanced Accuracy, 0.831 AUC Score
AnyScale Ranking Test for Hallucinations - 86.6 % Accuracy

Usage

The model can be used like this:

from sentence_transformers import CrossEncoder
model = CrossEncoder('vectara/hallucination_evaluation_model')
model.predict([
    ["A man walks into a bar and buys a drink", "A bloke swigs alcohol at a pub"],
    ["A person on a horse jumps over a broken down airplane.", "A person is at a diner, ordering an omelette."],
    ["A person on a horse jumps over a broken down airplane.", "A person is outdoors, on a horse."],
    ["A boy is jumping on skateboard in the middle of a red bridge.", "The boy skates down the sidewalk on a blue bridge"],
    ["A man with blond-hair, and a brown shirt drinking out of a public water fountain.", "A blond drinking water in public."],
    ["A man with blond-hair, and a brown shirt drinking out of a public water fountain.", "A blond man wearing a brown shirt is reading a book."],    
])

This returns a numpy array:

array([6.1051625e-01, 4.7493601e-04, 9.9639291e-01, 2.1221593e-04, 9.9599433e-01, 1.4126947e-03], dtype=float32)

Usage with Transformers AutoModel

You can use the model also directly with Transformers library (without SentenceTransformers library):

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

model = AutoModelForSequenceClassification.from_pretrained('vectara/hallucination_evaluation_model')
tokenizer = AutoTokenizer.from_pretrained('vectara/hallucination_evaluation_model')

pairs = [
    ["A man walks into a bar and buys a drink", "A bloke swigs alcohol at a pub"],
    ["A person on a horse jumps over a broken down airplane.", "A person is at a diner, ordering an omelette."],
    ["A person on a horse jumps over a broken down airplane.", "A person is outdoors, on a horse."],
    ["A boy is jumping on skateboard in the middle of a red bridge.", "The boy skates down the sidewalk on a blue bridge"],
    ["A man with blond-hair, and a brown shirt drinking out of a public water fountain.", "A blond drinking water in public."],
    ["A man with blond-hair, and a brown shirt drinking out of a public water fountain.", "A blond man wearing a brown shirt is reading a book."],    
]

inputs = tokenizer.batch_encode_plus(pairs, return_tensors='pt', padding=True)

model.eval()
with torch.no_grad():
    outputs = model(**inputs)
    logits = outputs.logits.cpu().detach().numpy()
    scores = 1 / (1 + np.exp(-logits)).flatten()

This returns a numpy array:

array([6.1051559e-01, 4.7493709e-04, 9.9639291e-01, 2.1221573e-04, 9.9599433e-01, 1.4127002e-03], dtype=float32)