GPT 1106
top-k1
{'exact_match': 30.41551246537396, 'f1': 40.33569098884983, 'tokens': 135.9382271468144, 'reader_fp': 0.2041551246537396, 'reader_fn': 0.03074792243767313, 'reader_precision': 0.5725058004640371, 'reader_recall': 0.8989071038251366}
top-k2
{'exact_match': 32.686980609418285, 'f1': 42.838019321398875, 'tokens': 272.75650969529084, 'reader_fp': 0.2598337950138504, 'reader_fn': 0.016897506925207757, 'reader_precision': 0.5439961108410306, 'reader_recall': 0.9483050847457627}
top-k4
{'exact_match': 33.29639889196676, 'f1': 44.4713305128544, 'tokens': 546.5484764542937, 'reader_fp': 0.33518005540166207, 'reader_fn': 0.009141274238227148, 'reader_precision': 0.49138293400588484, 'reader_recall': 0.9725457570715474}
top-k8
{'exact_match': 33.51800554016621, 'f1': 45.57744056497534, 'tokens': 1094.1670360110804, 'reader_fp': 0.39806094182825486, 'reader_fn': 0.004155124653739612, 'reader_precision': 0.4540273556231003, 'reader_recall': 0.987603305785124}
top-k16
{'exact_match': 33.961218836565095, 'f1': 46.26347198507874, 'tokens': 2189.9761772853185, 'reader_fp': 0.44930747922437675, 'reader_fn': 0.003601108033240997, 'reader_precision': 0.42786596119929454, 'reader_recall': 0.9893964110929854}
top-k32
{'exact_match': 34.51523545706371, 'f1': 46.73193531976337, 'tokens': 4382.209972299169, 'reader_fp': 0.489196675900277, 'reader_fn': 0.0019390581717451524, 'reader_precision': 0.4123128119800333, 'reader_recall': 0.9943820224719101}
4 → 1 1106
{'exact_match': 34.986149584487535, 'f1': 45.99523120960652, 'tokens': 121.31440443213296, 'reader_fp': 0.23988919667590028, 'reader_fn': 0.00775623268698061, 'reader_precision': 0.5878153260352214, 'reader_recall': 0.9778305621536025, 'sum_fp': 0.014404432132963989, 'sum_fn': 0.09889196675900278, 'sum_precision': 0.9752498810090433, 'sum_recall': 0.8516209476309227}
Seonglae Cho