Unverified50% confidenceFactExact time
非辛方法如标准RK4局部精度更高，但能量误差可能随时间单调增长
1
Sources
50%
Confidence
Long-term
Relevance
7/16/2026
First Seen
Sources
Related Claims
Unverified标准四阶龙格-库塔法（RK4）是耗散型积分器，能量误差可能随时间单调增长导致长时间尺度上的物理偏差84% similarUnverified低精度引入的数值扰动会在RL的多轮迭代中通过策略梯度的乘积形式被放大，可能导致训练崩溃或模型性能退化66% similarUnverifiedPost-training quantization solutions such as GPTQ and AWQ can control performance loss to within 1–2% at INT4 precision65% similarVerified4-bit quantization techniques (such as GPTQ, AWQ, and the GGUF format) can reduce model VRAM requirements to 1/4 of the original, but come with a certain degree of accuracy loss65% similarUnverifiedGPT-3研究中记录了「涌现能力（Emergent Abilities）」现象：当模型参数规模超过某个临界阈值，性能出现非线性跃升63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/533555API
curl https://kongchang.com/api/v1/knowledge/claims/533555MCP
get_claim(id=533555)