Online Self-Weighted Fine-Tuning
尚未评估计划受阻研究发现claim-main-table1-qwen3-4b

OSW-FT (K=2) on Qwen3-4B-Base achieves 13.33% Pass@1 and 36.67% Pass@16 on AIME 24, 16.25% Pass@1 and 36.67% Pass@16 on AIME 25, 41.42% Pass@1 and 74.42% Pass@16 on AMC 22, 45.11% Pass@1 and 78.26% Pass@16 on AMC 23, 35.00% Pass@1 and 75.56% Pass@16 on AMC 24, 79.00% Pass@1 on MATH-500, and 32.32% Pass@1 on GPQA-Diamond.

来源:source-paper:Table 1, Section 4.2, page 5

报告指标与观测值

pass_at_1

m-qwen3-4b-osw-aime24-p1

论文报告 13.33 percentage_points

实际观测 — percentage_points

pass_at_16

m-qwen3-4b-osw-aime24-p16

论文报告 36.67 percentage_points

实际观测 — percentage_points

pass_at_1

m-qwen3-4b-osw-aime25-p1

论文报告 16.25 percentage_points

实际观测 — percentage_points

pass_at_16

m-qwen3-4b-osw-aime25-p16

论文报告 36.67 percentage_points

实际观测 — percentage_points

pass_at_1

m-qwen3-4b-osw-math500-p1

论文报告 79 percentage_points

实际观测 — percentage_points

pass_at_1

m-qwen3-4b-osw-gpqad-p1

论文报告 32.32 percentage_points

实际观测 — percentage_points

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。