What does The Limits of Guardrail Compression conclude, in numbers, about “The Guardrail Source: Eight Kinds of Draft, Three Verdicts and the Counting Bound on Rules”?

Answer: Chapter 1 builds an illustrative source of eight kinds of draft and three verdicts and the counting bound: three rules keep 0.960 of the verdict information, two only 0.629.

Scope of the numbers: Chapter 1 builds an illustrative source of eight kinds of draft and three verdicts and the counting bound: three rules keep 0.960 of the verdict information, two only 0.629. Chapter 4 shows that the critical exchange rate is the reciprocal of the squared maximal correlation, 0.6844 and 1.4610 on the illustrative source while information rises continuously as rules split.

中文版

Data table

The numbers below are printed verbatim by the chapter's Python program and can be reproduced by running it. standard-library Python onlyprinted values that match the answer: 0.629, 0.960

partitions enumerated: 4140
klog2 kbest I(T;Y) shareaccuracyI(X;T)
10.0000.00000.0000.4580.000
21.0000.47370.6290.6270.971
31.5850.72340.9600.8051.559
42.0000.74110.9840.8051.812
52.3220.74830.9930.8052.075
62.5850.75220.9990.8052.310
72.8070.75290.9990.8052.443
83.0000.75341.0000.8052.552
fewest rules keeping 95% of I(X;Y): 3
best 3-rule grouping: [[0], [1, 2, 3], [4, 5, 6, 7]]
Show the full program
# Chapter 1. p12: every way to merge the eight draft kinds into k rules
# (all 4140 set partitions), the best decision information and accuracy
# each rule count can keep, and the counting bound I(X;T) <= log2 k.
# Source: the illustrative guardrail source of p11.
import math

PX = [0.40, 0.18, 0.10, 0.07, 0.10, 0.06, 0.05, 0.04]
PYX = [[0.92, 0.06, 0.02], [0.25, 0.70, 0.05], [0.30, 0.55, 0.15],
       [0.10, 0.60, 0.30], [0.05, 0.15, 0.80], [0.03, 0.12, 0.85],
       [0.02, 0.08, 0.90], [0.01, 0.04, 0.95]]


def entropy(p):
    return 0.0 - math.fsum(q * math.log2(q) for q in p if q > 0)


def mutual(pxy):
    px = [math.fsum(r) for r in pxy]
    py = [math.fsum(c) for c in zip(*pxy)]
    return math.fsum(v * math.log2(v / (px[i] * py[j]))
                     for i, r in enumerate(pxy)
                     for j, v in enumerate(r) if v > 0)


def partitions(items):
    if not items:
        yield []
        return
    first, rest = items[0], items[1:]
    for part in partitions(rest):
        for i in range(len(part)):
            yield part[:i] + [[first] + part[i]] + part[i + 1:]
        yield [[first]] + part


def merged(groups):
    return [[math.fsum(PX[x] * PYX[x][y] for x in g) for y in range(3)]
            for g in groups]


full = mutual([[a * b for b in r] for a, r in zip(PX, PYX)])
best = {}
count = 0
for part in partitions(list(range(8))):
    count += 1
    k = len(part)
    pty = merged(part)
    ity = mutual(pty)
    acc = math.fsum(max(r) for r in pty)
    ixt = entropy([math.fsum(r) for r in pty])
    old = best.get(k)
    if old is None or ity > old[0] + 1e-12:
        best[k] = (ity, acc, ixt, part)
print("partitions enumerated: %d" % count)
print(" k  log2 k  best I(T;Y) share  accuracy  I(X;T)")
for k in range(1, 9):
    ity, acc, ixt, part = best[k]
    print("%2d  %6.3f    %7.4f  %5.3f   %6.3f  %6.3f"
          % (k, math.log2(k), ity, ity / full, acc, ixt))
need = min(k for k in best if best[k][0] >= 0.95 * full)
print("fewest rules keeping 95%% of I(X;Y): %d" % need)
print("best 3-rule grouping: %s" % best[3][3])

Single source and citation

Fan, Y. (2026). The Information Bottleneck and Rate-Distortion Limits: How Far a Bilingual Monograph Can Be Compressed into Guardrails before Discriminative Power Falls Away (Version 1.0.0), Chapter 1: The Guardrail Source: Eight Kinds of Draft, Three Verdicts and the Counting Bound on Rules. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-250 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026250,
  author    = {Fan, Yafu},
  title     = {The Information Bottleneck and Rate-Distortion Limits: How Far a Bilingual Monograph Can Be Compressed into Guardrails before Discriminative Power Falls Away},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book