Unverified50% confidenceFactExact time
IDM uses measures such as KL divergence to minimize the difference between the student's reconstructed item distribution and the teacher's distribution.
1
Sources
50%
Confidence
Long-term
Relevance
10/26/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe distillation loss often uses KL divergence to measure the difference between the student distribution and the teacher distribution.75% 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 SD61% similarUnverified在标准投影方案下，教师分布的熵从2.09 bits骤降至0.32 bits，仅约15%的教师信息得以保留60% similarUnverifiedBy replacing the schema in the prompt with examples and normalizing enum values to lowercase before validation (using AI SDK's experimental_repairText), the pass rate of the 1B model can be improved from 1/10 to 6/10.59% similarUnverifiedCompared to the original baseline model, the student model after joint distillation and quantization achieves a 69% reduction in size.58% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1098103API
curl https://kongchang.com/api/v1/knowledge/claims/1098103MCP
get_claim(id=1098103)