Unverified50% confidenceFactExact time
The cache is stored in 4-bit format with one 8-bit scale shared per 16 channels, quantized after rotary positional encoding.
1
Sources
50%
Confidence
Long-term
Relevance
9/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedReal mode uses 16-bit registers and a segmented memory model, calculating a 20-bit physical address by shifting the 16-bit segment register left by 4 bits and adding a 16-bit offset, with an addressable space of only 1MB.74% similarUnverified位置编码外推方案可将模型有效上下文从最初的4K、8K扩展至128K乃至百万级token70% similarUnverifiedThe Mersenne Twister algorithm uses a state array of 624 32-bit integers, occupying approximately 2.5 KB of memory in total.69% similarUnverifieddecode阶段的生成速度近似等于内存带宽÷模型大小，例如在内存带宽273GB/s的M4 Pro上运行4GB的Q4量化7B模型理论最大生成速度约68 tokens/s69% similarUnverifiedext4 introduced the extent data structure, which describes contiguous regions using a starting block number plus a count of consecutive blocks; a single extent can describe up to 128 MB of contiguous space at a 4 KB block size.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/923872API
curl https://kongchang.com/api/v1/knowledge/claims/923872MCP
get_claim(id=923872)