Unverified50% confidenceBenchmarkExact time
The paper's training diagnostics show that ATI-OPD brings three observable improvements: fewer divergences between teacher and student, a decrease in the number of gradient spikes, and a reduction in the number of extremely long responses.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/30/2026
First Seen
Valid until: 1/28/2027
Sources
Related Entities
Related Claims
UnverifiedThe paper evaluates ATI-OPD on five mathematical reasoning benchmarks and two teacher-student model pairs, and the results show it improves the student's average accuracy in standalone reasoning compared with standard OPD, with gains covering multiple benchmarks and model pairs.74% similarVerifiedThe pre-training validation loss across the three generations of models shows a steady downward trend: 2.87 → 2.78 → 2.59.63% similarUnverified根据约翰·哈蒂对80,000余项教育研究的元分析《可见的学习》，教育干预的平均效应量约为0.4标准差61% similarUnverifiedUnder the multi-pass training mode, overall submission calibration error decreases by 63%–68%.61% similarUnverified哈蒂的元分析显示教育干预的平均效应量约为0.4 SD，其中反馈为0.73 SD、直接教学法0.60 SD、教师-学生关系0.72 SD、课堂规模缩减0.21 SD、开放式学习空间0.01 SD60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1114717API
curl https://kongchang.com/api/v1/knowledge/claims/1114717MCP
get_claim(id=1114717)