Unverified50% confidenceFactExact time
The paper proves that the leading term k/N is necessary and cannot be further removed or reduced, indicating that k is an essential parameter for characterizing the risk of compression learning.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTop-K的K值存在经典权衡：K太小会漏掉关键信息，K太大会引入噪声导致大模型答案跑偏70% similarUnverifiedThe choice of K value in Top-K routing affects the trade-off between computational efficiency and expressive capacity — a smaller K leads to faster inference but may sacrifice expressiveness, and in practice K is typically set to 1 or 2.69% similarUnverifiedThe complexity measure k is defined as the cardinality of the compression set plus the number of training samples on which the reconstructed hypothesis performs poorly.67% similarUnverified该测试采用KL散度（KLD）作为客观量化指标，衡量量化后输出分布与BF16参考分布的偏离程度，KLD越低说明量化造成的信息损失越小65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1001354API
curl https://kongchang.com/api/v1/knowledge/claims/1001354MCP
get_claim(id=1001354)