ListOps: 36.37Text: 64.27Retrieval: 57.46Image: 42.44
Papers with Code · CC BY-SA 4.0
Long-range modeling
A new task for testing the long-sequence modeling capabilities and efficiency of language models.
Image credit: SCROLLS: Standardized CompaRison Over Long Language Sequences
논문
GovRep: 61.1 / 32.3 / 33.7SumScr: 35.8 / 9.6 / 21.1QMSum: 34.9 / 11.8 / 23.5Qspr: 53.1
GovRep: 56.2 / 26.6 / 28.8SumScr: 24.2 / 4.5 / 15.4QMSum: 25.1 / 6.7 / 18.8Qspr: 26.6
GovRep: 60.3 / 30.0 / 31.5SumScr: 35.7 / 9.1 / 20.6QMSum: 33.2 / 9.6 / 21.6
GovRep: 53.6 / 26.1 / 28.8SumScr: 32.9 / 7.8 / 19.4QMSum: 31.1 / 8.5 / 20.4Qspr: 37.6
GovRep: 59.4 / 29.8 / 30.8SumScr: 37.7 / 10.2 / 21.5QMSum: 35.1 / 11.0 / 22.0Qspr: 48.7
GovRep: 61.3/32.2/33.8SumScr: 36.4/10.2/21.7QMSum: 36.2/12.9/24.3Qspr: 53.9
ListOps: 60.38Text: 86.44Retrieval: 83.41Image: 61.02
GovRep: 57.5 / 26.3 / 27.4SumScr: 35.2 / 8.7 / 19.4QMSum: 34.2 / 11.0 / 22.0Qspr: 46.9
ListOps: 59.60Text: 86.82Retrieval: 90.90Image: 88.65
ListOps: 62.15Text: 89.31Retrieval: 91.4Image: 88