1{2 "document_id": "lb_qasper_34e4bb298d4e2c19c2fce8670a00196cddd9f361eeb09a9a",3 "family": "public_qasper",4 "split": "dev",5 "document_hash": "964d476624da8b128caa47d71c9c02eb14daae7f3eded545935312e44377ae12",6 "questions": [7 {8 "question_id": "q00",9 "qtype": "qasper",10 "question": "How does Gaussian-masked directional multi-head attention works?",11 "answer": "pays attentions to adjacent characters and casts a localness relationship between the characters as a fixed Gaussian weight assuming the weight relies on the distance between characters",12 "aliases": [13 "Gaussian-masked directional attention can be described as a function to map queries and key-value pairs to the representation of input, Gaussian-masked directional attention expects to pay attention to the adjacent characters of each positions and cast the localness relationship between characters as a fix Gaussian weight for attention, Gaussian weight only relys on the distance between characters"14 ],15 "answer_type": "qa_f1",16 "tolerance": 0.0,17 "supporting_fact_ids": [],18 "answer_position_frac": -1.019 }20 ],21 "facts": []22}