Automatically detect hallucinations in a BART summarization model from its attentions and decoding probabilities - View it on GitHub