我们研究了和 $ A = \sum_{i=1}^N x_i $ 的全分布 $ P_{N}(A) $，其中 $ x_1, \dots, x_N $ 是 $ N \gg 1 $ 个独立同分布的随机变量，每个变量都从给定的分布 $ p(x) $ 中采样，且其在 $ x \to \infty $ 时具有次指数尾部。我们考虑两种特殊情况：(I) 单侧拉伸指数分布 $ p(x) \propto e^{-x^\alpha} $，其中 $ 0 < x < \infty $；(II) 双侧拉伸指数分布 $ p(x) \propto e^{-|x|^\alpha} $，其中 $ -\infty < x < \infty $。假设 $ 0 < \alpha < 1 $（在两种情况中均成立）。根据已知定理可直接得出：对于这两种情况，(i) $ \Delta A = A - \langle A \rangle $ 的典型涨落由中心极限定理描述；(ii) 尾部 $ A \to \infty $ 由大跳跃原理描述为 $ P_{N}(A) \simeq N p(A) $；(iii) 在这两个区域之间存在一个非平凡的中间区域，表现出反常标度 $ P_{N}(A) \sim e^{-N^\beta f(\Delta A / N^\gamma)} $，其中反常指数 $ \beta, \gamma \in (0,1) $ 和大偏差函数 $ f(y) $ 均为精确已知。在实际应用中，尽管这些理论预测在区域 (i) 和 (ii) 中表现良好，但在中间区域 (ii)，尤其是当 $ N $ 达到 $ 10^4 $ 时，误差往往达到几个数量级。我们计算了中间区域理论预测的次领阶修正。结果表明，对于 $ 0 < \alpha < \alpha_c $，这些修正按 $ N $ 的幂律缩放，而对于 $ \alpha_c < \alpha < 1 $，它们按拉伸指数缩放，阈值为 $ \alpha_c = 1/2 $（情况 I）和 $ \alpha_c = 2/3 $（情况 II）。这两种情况之间的差异是由于仅在后者中满足镜像对称性 $ p(x) = p(-x) $ 所致。

We study the full distribution $P_{N}\left(A\right)$ of sums $A = \sum_{i=1}^N$ where $x_1, \dots, x_N$ are $N \gg 1$ independent and identically distributed random variables each sampled from a given distribution $p(x)$ with a subexponential $x \to \infty$ tail. We consider two particular cases: (I) the one-sided stretched exponential distribution $p(x) \propto e^{-x^\alpha}$ where $0 < x < \infty$, (II) the two-sided stretched exponential distribution $p(x) \propto e^{-|x|^\alpha}$ where $-\infty < x < \infty$. We assume $0 < \alpha < 1$ (in both cases). As follows immediately from known theorems, for both cases (i) typical fluctuations of $\Delta A = A - \left\langle A\right\rangle $ are described by the central-limit theorem, (ii) the tail $A \to \infty$ is described by the big-jump principle $P_{N}\left(A\right) \simeq N p\left(A\right)$, and (iii) in between these two regimes there is a nontrivial intermediate regime which displays anomalous scaling $P_{N}\left(A\right) \sim e^{-N^\beta f(\Delta A/N^\gamma)}$ with anomalous exponents $\beta,\gamma \in (0,1)$ and large-deviation function $f(y)$ that are all exactly known. In practice, although these theoretical predictions of $P_{N}\left(A\right)$ work very well in regimes (i) and (ii), they often perform quite poorly in the intermediate regime (ii), with errors of several orders of magnitude for $N$ as large as $10^4$. We calculate subleading order corrections to the theoretical predictions in the intermediate regime. We find that for $0 < \alpha < \alpha_c$, these corrections scale as power laws in $N$, while for $\alpha_c < \alpha < 1$ they scale as stretched exponentials, where the threshold value is $\alpha_c = 1/2$ in case (I) and $\alpha_c = 2/3$ in case (II). This difference between the two cases is a result of the mirror symmetry $p(x) = p(-x)$ which holds only in the latter case.