Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality
尚未评估已列入计划研究发现claim-downstream-zero-shot-370m
Mamba-2-370M trained on 300B tokens of the Pile achieves zero-shot downstream task performance of 55.8% on LAMBADA, 46.9% on HellaSwag, 70.5% on PIQA, 54.9% on Arc-Easy, 26.9% on Arc-Challenge, 55.7% on WinoGrande, 32.4% on OpenbookQA, and 49.0% average accuracy across tasks.
来源：source-paper:Table 10, page 52
报告指标与观测值
lambada_acc
rm-370m-lambada-acc
论文报告 55.8 percent
实际观测 — percent
hellaswag_acc
rm-370m-hellaswag-acc
论文报告 46.9 percent
实际观测 — percent
piqa_acc
rm-370m-piqa-acc
论文报告 70.5 percent
实际观测 — percent
arc_easy_acc
rm-370m-arc-e-acc
论文报告 54.9 percent
实际观测 — percent
arc_challenge_acc
rm-370m-arc-c-acc
论文报告 26.9 percent
实际观测 — percent
winogrande_acc
rm-370m-winogrande-acc
论文报告 55.7 percent
实际观测 — percent
openbookqa_acc
rm-370m-openbookqa-acc
论文报告 32.4 percent
实际观测 — percent
average_acc
rm-370m-avg-acc
论文报告 49 percent
实际观测 — percent
Assessment（0）
这条 Claim 暂无已发布的不可变 Assessment。
关联运行（0）
这条 Claim 暂无关联执行。