关于「伏笔的遗忘衰减:最优回收时点、中途提醒与并行线索上限」,《长篇叙事的因果图、悬念熵与伏笔回收》给出的定量结论是什么?
结论:第五章给出伏笔最优回收时点的解析式,一次位置合适的轻提醒把满足感从 0.018 提到 0.678,同时依赖 6 条伏笔的高潮被完整理解的概率只有 0.531。
数据口径:书中的连载、读者与参数均为示意模型;全部计算由书中程序生成;本书不是法律、财务或投资建议,并坚持一条红线:不得以虚假悬念骗取点击。 第一、二章用卡恩拓扑排序与按键分组的版本检查找出时间悖论与同一属性的两个值,并说明跨卷吃书藏在推得出的先后里:示意连载第六卷的一条边与写在第 2 卷到第 6 卷之间的 10 条边相冲突。
对照数据表
下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 0.018, 0.678
| payoff at chapter 60, initial stability 15 chapters | |||
| plan | satisfaction | recall@P | stability |
| no reminder | 0.018 | 0.018 | 15.0 |
| one at ch 30 | 0.417 | 0.419 | 34.5 |
| one at ch 57 | 0.586 | 0.922 | 37.0 |
| best one: ch 51 | 0.678 | 0.783 | 36.7 |
| two even: ch 20, 40 | 0.674 | 0.689 | 53.8 |
| best two: ch 19, 49 | 0.752 | 0.833 | 60.1 |
| best three: ch 10, 25, 49 | 0.772 | 0.856 | 70.6 |
| marginal gain of reminders 1, 2, 3: 0.659, 0.074, 0.020 |
查看完整验算程序
# Chapter 5. p52: reminders for a long-range foreshadowing.
# A clue planted at chapter 0 pays off at chapter P = 60. Recall decays as
# exp(-age / S) from the last exposure. A light reminder at chapter r
# resets recall to 1 and, as in the spacing effect, raises stability more
# the more had been forgotten: S <- S (1 + G (1 - R(r))). A reminder d
# chapters before the payoff gives the payoff away with weight
# W exp(-d / DELTA). Satisfaction = appetite A(P) x recall at P x
# (1 - spoil). Exhaustive search over one, two and three positions.
# All parameters are illustrative.
import itertools
import math
P, S0, TAU, G, W, DELTA = 60, 15.0, 8.0, 1.5, 0.6, 6.0
def satisfaction(reminders):
s, last = S0, 0
for r in sorted(reminders):
rec = math.exp(-(r - last) / s)
s *= 1 + G * (1 - rec)
last = r
recall = math.exp(-(P - last) / s)
spoil = math.fsum(W * math.exp(-(P - r) / DELTA) for r in reminders)
appetite = 1 - math.exp(-P / TAU)
return appetite * recall * max(0.0, 1 - spoil), recall, s
base = satisfaction(())
print("payoff at chapter {}, initial stability {:.0f} chapters".format(
P, S0))
print("plan satisfaction recall@P stability")
print("{:<25} {:>12.3f} {:>8.3f} {:>9.1f}".format("no reminder", *base))
rows = [("one at ch 30", (30,)), ("one at ch 57", (57,))]
best1 = max(range(1, P), key=lambda r: satisfaction((r,))[0])
rows.append(("best one: ch {}".format(best1), (best1,)))
rows.append(("two even: ch 20, 40", (20, 40)))
best2 = max(itertools.combinations(range(1, P), 2),
key=lambda rr: satisfaction(rr)[0])
rows.append(("best two: ch {}, {}".format(*best2), best2))
for label, rr in rows:
print("{:<25} {:>12.3f} {:>8.3f} {:>9.1f}".format(
label, *satisfaction(rr)))
best3 = max(itertools.combinations(range(1, P), 3),
key=lambda rr: satisfaction(rr)[0])
print("{:<25} {:>12.3f} {:>8.3f} {:>9.1f}".format(
"best three: ch {}, {}, {}".format(*best3), *satisfaction(best3)))
m = [satisfaction(x)[0] for x in ((), (best1,), best2, best3)]
print("marginal gain of reminders 1, 2, 3: {:.3f}, {:.3f}, {:.3f}".format(
m[1] - m[0], m[2] - m[1], m[3] - m[2]))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《长篇叙事因果有向无环图校验、章节悬念信息熵锯齿调度与伏笔遗忘衰减回收》第 5 章「伏笔的遗忘衰减:最优回收时点、中途提醒与并行线索上限」(版本 1.0.0). 5D Library. DOI 待发布
DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-178 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026178,
author = {Fan, Yafu},
title = {Causal-DAG Validation of Long-Form Narrative, Sawtooth Scheduling of Chapter Suspense Entropy and Forgetting-Aware Foreshadowing Recovery},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}