Unverified60% confidenceFactExact time
Ternary quantization means each weight has only three possible value states, typically -1, 0, and +1.
2
Sources
60%
Confidence
Long-term
Relevance
8/5/2026
First Seen
Sources
Related Claims
Unverified4-bit量化只能表示16个离散值，1-bit只有2个，三元量化（-1/0/+1）只有3个可能值72% similarUnverifiedTernary quantization theoretically requires only approximately 1.58 bits (log₂3) per weight for representation.70% similarUnverifiedq4_0使用对称量化（零点固定为0），q4_1使用非对称量化（零点可浮动），后者在数值分布不对称时更精确但需额外存储零点参数69% similarUnverifiedThe 1st term is 1^1+0=1, the 2nd term is 2^2+1=5, the 3rd term is 3^3+2=29, and the 4th term is 4^4+3=259.67% similarUnverifiedTernary weights use absmean quantization: γ is computed as the mean of the absolute values of the weights, |w| <= γ/2 becomes 0, w > γ/2 becomes +1, and w < -γ/2 becomes -1; activations use symmetric absmax quantization.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/692439API
curl https://kongchang.com/api/v1/knowledge/claims/692439MCP
get_claim(id=692439)