title 符號、神經與代數不是三個世界:表示制度、共同結構與三重計算體系
english_title Symbolic, Neural, and Algebraic Are Not Three Worlds: Representation Regimes, Shared Structure, and a Triple Computational System
series 理論域擴張、理解解析度與共同底空間
series_english Theory-Domain Expansion, Comprehension Resolution, and Shared-Substrate Theory
series_id TCRS
paper_id TCRS-08
author Neo.K
organization EveMissLab
version 0.1.0
status Research Draft / Series Paper
date 2026-10-06
language zh-TW
canonical_source UTF-8 Markdown
math_source_rule inline math only with $...$ ; display math only with $$...$$
符號、神經與代數不是三個世界
表示制度、共同結構與三重計算體系
Symbolic, Neural, and Algebraic Are Not Three Worlds: Representation Regimes, Shared Structure, and a Triple Computational System
系列: 理論域擴張、理解解析度與共同底空間(TCRS),Paper 08作者: Neo.K機構: EveMissLab版本: v0.1.0日期: 2026-10-06
公司: 一言諾科技有限公司
摘要
AI 歷史常被粗略分成:
Symbolic v s Neural . \boxed{
\text{Symbolic}
\quad vs\quad
\text{Neural}.
} Symbolic v s Neural .
近年又可以再加入一類更底層的 structural / algebraic approaches:
Symbolic v s Neural v s Algebraic / Structural . \boxed{
\text{Symbolic}
\quad vs\quad
\text{Neural}
\quad vs\quad
\text{Algebraic / Structural}.
} Symbolic v s Neural v s Algebraic / Structural .
但本文主張,這三者不應首先被理解為互斥的「智能本體」。
更合理的理解是:
three representation regimes with different computational advantages . \boxed{
\text{three representation regimes with different computational advantages}.
} three representation regimes with different computational advantages .
本文將其抽象為:
S = Symbolic Regime , \boxed{
S
=
\text{Symbolic Regime},
} S = Symbolic Regime ,
N = Neural Regime , \boxed{
N
=
\text{Neural Regime},
} N = Neural Regime ,
A = Algebraic / Structural Regime . \boxed{
A
=
\text{Algebraic / Structural Regime}.
} A = Algebraic / Structural Regime .
三者分別具有不同核心能力:
S : Identity + Addressability + Explicit Rule + Verification , \boxed{
S:
\text{Identity}
+
\text{Addressability}
+
\text{Explicit Rule}
+
\text{Verification},
} S : Identity + Addressability + Explicit Rule + Verification ,
N : Discovery + Adaptation + Distributed Representation + Approximation , \boxed{
N:
\text{Discovery}
+
\text{Adaptation}
+
\text{Distributed Representation}
+
\text{Approximation},
} N : Discovery + Adaptation + Distributed Representation + Approximation ,
A : Invariant + Composition + Transformation + Equivalence . \boxed{
A:
\text{Invariant}
+
\text{Composition}
+
\text{Transformation}
+
\text{Equivalence}.
} A : Invariant + Composition + Transformation + Equivalence .
傳統 neuro-symbolic AI 常被理解為:
N + S . \boxed{
N+S.
} N + S .
但本文提出,真正缺少的可能是一個中介結構層:
N ↔ A ↔ S . \boxed{
N
\leftrightarrow
A
\leftrightarrow
S.
} N ↔ A ↔ S .
其中 Algebraic / Structural layer 不只是「又一種符號」,而是負責描述:
哪些 transformation 保持不變;
哪些 representation 可以視為等價;
哪些局部結構可組合;
哪些高維狀態可 canonicalize;
哪些 symbolic objects 可以編譯回 neural substrate。
因此本文將未來可能的核心問題重新表述為:
How can the same structure survive across different computational coordinate systems? \boxed{
\text{How can the same structure survive across different computational coordinate systems?}
} How can the same structure survive across different computational coordinate systems?
而不是:
Which school is the true form of intelligence? \boxed{
\text{Which school is the true form of intelligence?}
} Which school is the true form of intelligence?
本文進一步提出:
Representation-Regime Separation;
Identity–Geometry–Invariant Triangle;
Addressability Gap;
Structural Mediation Layer;
Bidirectional Representation Compilation;
Representation-Preserving Translation;
Canonicalization Witness;
Structural Conservation;
Regime-Relative Completeness;
Triple-Regime Architecture。
本文指出,Symbolic、Neural、Algebraic 都可能在自己的 domain 內形成類完備感,並發生 TCRS 前文所述的:
Mutual Theoretical Absorption . \boxed{
\text{Mutual Theoretical Absorption}.
} Mutual Theoretical Absorption .
Symbolic 可能說:
neural 最後不就是規則與 relation。
Neural 可能說:
symbolic 最後不就是可學 pattern。
Algebraic 可能說:
兩者最後不都是 transformation structure。
而真正的 Green Model 不是讓其中一派勝出,而是建立:
representation-independent shared structure \boxed{
\text{representation-independent shared structure}
} representation-independent shared structure
並允許不同 regime 在不同 bottleneck 下承擔不同角色。
本文的核心命題為:
Symbolic, neural, and algebraic systems may be better understood as different coordinate systems over partially shared computational structure . \boxed{
\text{Symbolic, neural, and algebraic systems may be better understood as different coordinate systems over partially shared computational structure}.
} Symbolic, neural, and algebraic systems may be better understood as different coordinate systems over partially shared computational structure .
關鍵詞: Symbolic AI、Neural AI、Algebraic AI、Structural AI、Representation Regime、Bidirectional Compilation、Canonicalization、Shared Structure、Neuro-Symbolic、AI Architecture
1. 問題起點:三派真的在研究三種不同智能嗎?
如果只看歷史敘事,很容易得到:
S v s N v s A . \boxed{
S
\quad vs\quad
N
\quad vs\quad
A.
} S v s N v s A .
2. 但這可能把表示形式誤認成智能本體
也就是:
Representation → Ontology . \boxed{
\text{Representation}
\rightarrow
\text{Ontology}.
} Representation → Ontology .
3. Symbolic 的世界
Symbolic 傾向使用:
named objects;
predicates;
rules;
proofs;
explicit relations。
4. Neural 的世界
Neural 傾向使用:
vectors;
matrices;
activations;
weights;
nonlinear transformations;
probabilistic distributions。
5. Algebraic / Structural 的世界
Algebraic 傾向使用:
groups;
rings;
fields;
categories;
operators;
invariants;
equivalence classes;
morphisms。
6. 三者語言非常不同
所以很容易形成:
Language Difference → Ontological Difference . \boxed{
\text{Language Difference}
\rightarrow
\text{Ontological Difference}.
} Language Difference → Ontological Difference .
7. 但這個推論不必成立
可能存在某個:
S ∗ \boxed{
\mathcal S^\ast
} S ∗
使:
S = Π S ( S ∗ ) , S=\Pi_S(\mathcal S^\ast), S = Π S ( S ∗ ) ,
N = Π N ( S ∗ ) , N=\Pi_N(\mathcal S^\ast), N = Π N ( S ∗ ) ,
A = Π A ( S ∗ ) . A=\Pi_A(\mathcal S^\ast). A = Π A ( S ∗ ) .
8. 這不代表三者完全等價
而是:
它們可能保留同一底層結構的不同部分。
9. Representation-Regime Separation
本文提出:
Representation Regime ≠ Intelligence Ontology . \boxed{
\text{Representation Regime}
\neq
\text{Intelligence Ontology}.
} Representation Regime = Intelligence Ontology .
10. Symbolic Regime
本文用:
S \boxed{
S
} S
表示 symbolic regime。
11. S 的第一核心:Identity
Symbolic 很擅長:
x = x . \boxed{
x=x.
} x = x .
12. Named Object
例如:
E n t i t y 17 \boxed{
Entity_{17}
} E n t i t y 17
就是同一個可追蹤 object。
13. 第二核心:Addressability
可以直接:
Access ( E n t i t y 17 ) . \boxed{
\operatorname{Access}(Entity_{17}).
} Access ( E n t i t y 17 ) .
14. 第三核心:Explicit Relation
R ( E n t i t y 17 , E n t i t y 23 ) . \boxed{
R(Entity_{17},Entity_{23}).
} R ( E n t i t y 17 , E n t i t y 23 ) .
15. 第四核心:Rule
P ( x ) ∧ Q ( x ) → R ( x ) . \boxed{
P(x)
\land
Q(x)
\rightarrow
R(x).
} P ( x ) ∧ Q ( x ) → R ( x ) .
16. 第五核心:Verification
如果 rule system well-defined,
可以:
trace → proof . \boxed{
\text{trace}
\rightarrow
\text{proof}.
} trace → proof .
17. Symbolic Strength Vector
本文暫寫:
Q ⃗ S = ( I , A , R , V ) . \boxed{
\vec Q_S
=
(
I,
A,
R,
V
).
} Q S = ( I , A , R , V ) .
其中:
I I I :identity;
A A A :addressability;
R R R :rule explicitness;
V V V :verifiability。
18. Neural Regime
本文用:
N \boxed{
N
} N
表示 neural regime。
19. N 的第一核心:Distributed Representation
一個 concept 未必存在於:
w i . w_i. w i .
而可能分散於:
Z ⊂ R d . \boxed{
Z
\subset
\mathbb R^d.
} Z ⊂ R d .
20. 第二核心:Adaptation
Neural 可以:
D + O t r a i n → Z . \boxed{
D
+
\mathcal O_{\mathrm{train}}
\rightarrow
Z.
} D + O train → Z .
21. 第三核心:Approximation
對複雜 function:
f f f
學得:
f ^ θ . \boxed{
\hat f_\theta.
} f ^ θ .
22. 第四核心:High-Dimensional Association
大量弱相關訊號可以在:
Z Z Z
中形成可用 geometry。
23. Neural Strength Vector
Q ⃗ N = ( D r , A d , A p , H ) . \boxed{
\vec Q_N
=
(
D_r,
A_d,
A_p,
H
).
} Q N = ( D r , A d , A p , H ) .
其中:
D r D_r D r :distributed representation;
A d A_d A d :adaptation;
A p A_p A p :approximation;
H H H :high-dimensional association。
24. Algebraic / Structural Regime
本文用:
A \boxed{
A
} A
表示 algebraic / structural regime。
25. A 的第一核心:Invariant
問:
what remains unchanged under transformation? \boxed{
\text{what remains unchanged under transformation?}
} what remains unchanged under transformation?
26. 第二核心:Composition
如果:
f : A → B , f:A\rightarrow B, f : A → B ,
g : B → C , g:B\rightarrow C, g : B → C ,
則:
g ∘ f : A → C . \boxed{
g\circ f:A\rightarrow C.
} g ∘ f : A → C .
27. 第三核心:Transformation
研究:
T : X → X ′ . \boxed{
T:X\rightarrow X'.
} T : X → X ′ .
本身的結構。
28. 第四核心:Equivalence
例如:
x ∼ y . \boxed{
x\sim y.
} x ∼ y .
雖表示不同,
但在某 structure 下等價。
29. Algebraic Strength Vector
Q ⃗ A = ( I v , C , T , E q ) . \boxed{
\vec Q_A
=
(
I_v,
C,
T,
E_q
).
} Q A = ( I v , C , T , E q ) .
30. 三者不是同一軸上的高低
不能簡化成:
S < N < A S<N<A S < N < A
或:
A < S < N . A<S<N. A < S < N .
31. 更合理是三個不同優勢軸
R = { S , N , A } . \boxed{
\mathcal R
=
\{S,N,A\}.
} R = { S , N , A } .
32. Identity–Geometry–Invariant Triangle
本文提出:
Identity ↔ Geometry ↔ Invariant . \boxed{
\text{Identity}
\leftrightarrow
\text{Geometry}
\leftrightarrow
\text{Invariant}.
} Identity ↔ Geometry ↔ Invariant .
33. Symbolic 偏 Identity
S → identity / explicit relation . \boxed{
S
\rightarrow
\text{identity / explicit relation}.
} S → identity / explicit relation .
34. Neural 偏 Geometry
N → distributed geometry / state field . \boxed{
N
\rightarrow
\text{distributed geometry / state field}.
} N → distributed geometry / state field .
35. Algebraic 偏 Invariant
A → transformation-invariant structure . \boxed{
A
\rightarrow
\text{transformation-invariant structure}.
} A → transformation-invariant structure .
36. 三者其實互相需要
Identity 沒有 geometry,
可能 rigid。
37. Geometry 沒有 identity
可能難以 address。
38. Invariant 沒有 concrete realization
可能缺少 learnable substrate。
39. 所以未來問題不是選一個
而是:
how to map among them . \boxed{
\text{how to map among them}.
} how to map among them .
40. Neuro-Symbolic 的基本問題
傳統想法:
N + S . \boxed{
N+S.
} N + S .
41. 但 N 與 S 之間存在表示裂縫
Neural state:
z ∈ R d . z\in\mathbb R^d. z ∈ R d .
Symbolic object:
s ∈ Σ ∗ s\in\Sigma^\ast s ∈ Σ ∗
或:
s ∈ K . s\in\mathcal K. s ∈ K .
42. 直接 mapping 很難
因為:
z ↮ s \boxed{
z
\not\leftrightarrow
s
} z ↔ s
通常不是一一對應。
43. 多對多
可能:
{ z i } ↔ { s j } . \boxed{
\{z_i\}
\leftrightarrow
\{s_j\}.
} { z i } ↔ { s j } .
44. 還有 superposition
一個 neural state 可能同時承載:
s 1 , s 2 , … , s k . \boxed{
s_1,s_2,\ldots,s_k.
} s 1 , s 2 , … , s k .
45. 也有 context dependence
同一:
z z z
在不同:
c c c
下可以不同解讀。
46. 所以純 N→S 解碼容易 lossy
N → S \boxed{
N
\rightarrow
S
} N → S
不一定保持:
geometry;
uncertainty;
latent relations。
47. Structural Mediation Hypothesis
本文提出:
N → A → S \boxed{
N
\rightarrow
A
\rightarrow
S
} N → A → S
可能比:
N → S N\rightarrow S N → S
更穩定。
48. A 的功能不是「再加一層符號」
而是先問:
在 neural representation 裡,哪些結構具有可保持性?
49. Structural Extraction
從:
Z Z Z
抽:
I Z \boxed{
\mathcal I_Z
} I Z
即 invariant structure。
50. 可能包含
equivalence class;
relational symmetry;
transformation orbit;
compositional operator;
conserved constraint。
51. 再 canonicalize
I Z → C Z . \boxed{
\mathcal I_Z
\rightarrow
C_Z.
} I Z → C Z .
52. Canonical Form
C Z \boxed{
C_Z
} C Z
是:
對同一結構選一個穩定、可定址的標準表示。
53. 這時再進 Symbolic
C Z → S Z . \boxed{
C_Z
\rightarrow
S_Z.
} C Z → S Z .
54. 整體:
N → A → S . \boxed{
N
\rightarrow
A
\rightarrow
S.
} N → A → S .
55. 反方向也重要
Symbolic specification:
S S S
可以先轉成:
A S \boxed{
A_S
} A S
即結構約束。
56. 再實現在 neural substrate
A S → N S . \boxed{
A_S
\rightarrow
N_S.
} A S → N S .
57. 所以:
S → A → N . \boxed{
S
\rightarrow
A
\rightarrow
N.
} S → A → N .
58. Bidirectional Representation Compilation
本文正式提出:
BRC = Bidirectional Representation Compilation . \boxed{
\text{BRC}
=
\text{Bidirectional Representation Compilation}.
} BRC = Bidirectional Representation Compilation .
59. 核心
N ↔ A ↔ S . \boxed{
N
\leftrightarrow
A
\leftrightarrow
S.
} N ↔ A ↔ S .
60. 這不是簡單 translation
而是:
structure-preserving compilation . \boxed{
\text{structure-preserving compilation}.
} structure-preserving compilation .
61. Representation-Preserving Translation
本文提出:
RPT . \boxed{
\text{RPT}.
} RPT .
要求:
I ( X ) ≈ I ( Φ ( X ) ) \boxed{
I(
X
)
\approx
I(
\Phi(X)
)
} I ( X ) ≈ I ( Φ ( X ))
對關鍵 invariant 成立。
62. Translation 不是只看 output match
兩個 representation:
X , Y X,Y X , Y
可能 output 一樣,
但 internal structure 完全不同。
63. 所以需要 structural witness
Γ Φ . \boxed{
\Gamma_{\Phi}.
} Γ Φ .
64. Canonicalization Witness
本文提出:
Γ C . \boxed{
\Gamma_C.
} Γ C .
表示:
為什麼一組不同表示可以被合法視為同一 canonical structure。
65. 沒有 witness 的 canonicalization
可能只是:
forced discretization . \boxed{
\text{forced discretization}.
} forced discretization .
66. Forced Discretization
Neural state:
z z z
被硬切成:
s . s. s .
但實際:
uncertainty \boxed{
\text{uncertainty}
} uncertainty
與:
superposition \boxed{
\text{superposition}
} superposition
被丟失。
67. 所以 symbolic layer 不應吞掉 neural residual
這又回到 TCRS-05。
68. N→S 也要保存 residual
可寫:
N → ( S , R N ) . \boxed{
N
\rightarrow
(
S,
R_N
).
} N → ( S , R N ) .
69. R_N 是 neural residual
R N \boxed{
R_N
} R N
表示:
尚不能被 canonical symbolic structure 完整表示的部分。
70. 這很重要
否則:
canonicalization → false completeness . \boxed{
\text{canonicalization}
\rightarrow
\text{false completeness}.
} canonicalization → false completeness .
71. Symbolic 也有 residual
反向:
S → ( N , R S ) . \boxed{
S
\rightarrow
(
N,
R_S
).
} S → ( N , R S ) .
72. 某些 symbolic constraint
可能很難自然嵌入:
N . N. N .
73. 所以雙向都不一定 lossless
L N → S > 0 \boxed{
L_{N\rightarrow S}>0
} L N → S > 0
L S → N > 0. \boxed{
L_{S\rightarrow N}>0.
} L S → N > 0.
74. Algebraic layer 的價值
就是把 loss 顯性化。
75. Structural Conservation
本文提出:
Structural Conservation . \boxed{
\text{Structural Conservation}.
} Structural Conservation .
76. 定義
跨 representation 時,不要求:
X = Y , X=Y, X = Y ,
而要求:
I ( X ) = I ( Y ) \boxed{
\mathcal I(X)
=
\mathcal I(Y)
} I ( X ) = I ( Y )
對選定 invariant 成立。
77. 這比 literal identity 更實際
因為:
N , S , A N,S,A N , S , A
本來就不是同一表示。
78. Representation Coordinate System
本文將每個 regime 視為一種:
computational coordinate system . \boxed{
\text{computational coordinate system}.
} computational coordinate system .
79. Neural Coordinate
偏:
continuous / distributed . \boxed{
\text{continuous / distributed}.
} continuous / distributed .
80. Symbolic Coordinate
偏:
discrete / named / explicit . \boxed{
\text{discrete / named / explicit}.
} discrete / named / explicit .
81. Algebraic Coordinate
偏:
relation / transformation / equivalence . \boxed{
\text{relation / transformation / equivalence}.
} relation / transformation / equivalence .
82. 問題變成:
what survives coordinate change? \boxed{
\text{what survives coordinate change?}
} what survives coordinate change?
83. 這是比「誰比較高級」更好的問題
因為它直接指向:
I ∗ . \boxed{
\mathcal I^\ast.
} I ∗ .
84. Representation-Independent Invariant
本文提出:
I ∗ \boxed{
\mathcal I^\ast
} I ∗
為:
在多個 representation regime 間仍可保持的核心 structure。
85. 這可能包括
causal relation;
equivalence;
order;
constraint;
symmetry;
transition structure。
86. 不是所有資訊都 invariant
重要。
87. 有些資訊只存在於某 regime
例如:
fine-grained geometry \boxed{
\text{fine-grained geometry}
} fine-grained geometry
可能是 neural-specific。
88. 有些則是 symbolic-specific
例如:
explicit identifier provenance . \boxed{
\text{explicit identifier provenance}.
} explicit identifier provenance .
89. 所以共同底空間不等於全資訊交集
而是:
shared invariant + regime-specific residual . \boxed{
\text{shared invariant}
+
\text{regime-specific residual}.
} shared invariant + regime-specific residual .
90. Regime-Relative Completeness
本文提出:
Regime-Relative Completeness . \boxed{
\text{Regime-Relative Completeness}.
} Regime-Relative Completeness .
91. Neural 可以在 perception domain 類完備
例如:
Q N ( D p ) ≈ 1. Q_N(D_p)\approx1. Q N ( D p ) ≈ 1.
92. Symbolic 可以在 formal verification 類完備
Q S ( D v ) ≈ 1. Q_S(D_v)\approx1. Q S ( D v ) ≈ 1.
93. Algebraic 可以在 structural equivalence 類完備
Q A ( D e ) ≈ 1. Q_A(D_e)\approx1. Q A ( D e ) ≈ 1.
94. 但:
Q N ( D p ) ≈ 1 ⇏ Q N ( Ω ) ≈ 1. \boxed{
Q_N(D_p)\approx1
\not\Rightarrow
Q_N(\Omega)\approx1.
} Q N ( D p ) ≈ 1 ⇒ Q N ( Ω ) ≈ 1.
95. 三派都可能犯「萬法皆我」
Symbolic:
所有 neural 最後都是 rules。
96. Neural:
所有 rules 最後都能被 network 學。
97. Algebraic:
所有系統最後都只是 algebraic structure。
98. 三句都可能部分正確
但:
representable ≠ exhausted . \boxed{
\text{representable}
\neq
\text{exhausted}.
} representable = exhausted .
99. Mutual Regime Absorption
本文提出:
MRA = Mutual Regime Absorption . \boxed{
\text{MRA}
=
\text{Mutual Regime Absorption}.
} MRA = Mutual Regime Absorption .
100. 三方各自都能說:
the others are special cases of me . \boxed{
\text{the others are special cases of me}.
} the others are special cases of me .
101. 這是 TCRS-05 的 AI 特例
即:
Mutual Theoretical Absorption . \boxed{
\text{Mutual Theoretical Absorption}.
} Mutual Theoretical Absorption .
102. 真正共同底空間要 de-privilege 三方
所以:
S ∗ \boxed{
S^\ast
} S ∗
不能預設:
symbol primitive;
vector primitive;
algebraic primitive;
哪一個必然最底層。
103. Shared Structural Substrate
本文暫定義:
S ∗ = ( I , T , C , R ) . \boxed{
\mathcal S^\ast
=
(
I,
T,
C,
R
).
} S ∗ = ( I , T , C , R ) .
其中:
I I I :invariants;
T T T :transformations;
C C C :constraints;
R R R :residual channels。
104. Symbolic Projection
Π S ( S ∗ ) . \boxed{
\Pi_S(
\mathcal S^\ast
).
} Π S ( S ∗ ) .
105. Neural Projection
Π N ( S ∗ ) . \boxed{
\Pi_N(
\mathcal S^\ast
).
} Π N ( S ∗ ) .
106. Algebraic Projection
Π A ( S ∗ ) . \boxed{
\Pi_A(
\mathcal S^\ast
).
} Π A ( S ∗ ) .
107. 不要求三者完全可逆
而要求:
explicitly measured reversibility . \boxed{
\text{explicitly measured reversibility}.
} explicitly measured reversibility .
108. Representation Round-Trip
例如:
N → A → S → A → N ^ . \boxed{
N
\rightarrow
A
\rightarrow
S
\rightarrow
A
\rightarrow
\hat N.
} N → A → S → A → N ^ .
109. Round-Trip Error
E N A S A N = d ( N , N ^ ) . \boxed{
E_{NASAN}
=
d(
N,\hat N
).
} E N A S A N = d ( N , N ^ ) .
110. 同樣:
E S A N A S \boxed{
E_{SANAS}
} E S A N A S
可以測反向。
111. 這比單向 accuracy 更有意義
因為可以暴露:
translation loss . \boxed{
\text{translation loss}.
} translation loss .
112. Addressability Gap
本文提出:
G A = Addressability Gap . \boxed{
G_A
=
\text{Addressability Gap}.
} G A = Addressability Gap .
113. 定義
Neural representation 擁有能力,但無法穩定指向其內部可編輯、可驗證結構的程度。
114. Symbolic 的核心優勢
就是:
G A ↓ . \boxed{
G_A\downarrow.
} G A ↓ .
115. Geometry Gap
反過來 Symbolic 可能有:
G G = Geometry Gap . \boxed{
G_G
=
\text{Geometry Gap}.
} G G = Geometry Gap .
116. 即:
難以自然表達高維連續、模糊、分布式相似性。
117. Structural Gap
如果只有 N+S,
中間可能仍缺:
G N S . \boxed{
G_{NS}.
} G N S .
118. 也就是:
如何判定哪些 neural states 應該對應哪些 symbolic structures?
119. Algebraic layer 可提供
equivalence criteria . \boxed{
\text{equivalence criteria}.
} equivalence criteria .
120. 這就是中介價值
不是因為 algebra 比另外兩者「更真」。
121. 而是因為它擅長:
cross-representation structure . \boxed{
\text{cross-representation structure}.
} cross-representation structure .
122. Triple-Regime Architecture
本文提出:
A S N A \boxed{
\mathcal A_{SNA}
} A S N A
123. 其最小結構
N ↔ A ↔ S . \boxed{
N
\leftrightarrow
A
\leftrightarrow
S.
} N ↔ A ↔ S .
124. Neural Layer
負責:
perception;
learning;
approximation;
high-dimensional association。
125. Algebraic Layer
負責:
invariant extraction;
canonicalization;
compositional relation;
coordinate translation。
126. Symbolic Layer
負責:
naming;
addressability;
explicit rule;
audit;
local repair。
127. 這不是固定分工
不同 task:
Q Q Q
會有不同:
w ⃗ ∗ ( Q ) = ( w N , w A , w S ) . \boxed{
\vec w^\ast(Q)
=
(
w_N,w_A,w_S
).
} w ∗ ( Q ) = ( w N , w A , w S ) .
128. Dynamic Regime Routing
本文提出:
Dynamic Regime Routing . \boxed{
\text{Dynamic Regime Routing}.
} Dynamic Regime Routing .
129. Perception
可能:
w N ≫ w A , w S . w_N\gg w_A,w_S. w N ≫ w A , w S .
130. Formal Proof
可能:
w S , w A ≫ w N . w_S,w_A\gg w_N. w S , w A ≫ w N .
131. World-Model Update
可能:
w N ≈ w A ≈ w S . w_N\approx w_A\approx w_S. w N ≈ w A ≈ w S .
132. 所以未來架構更像 ecology
而不是:
single universal representation . \boxed{
\text{single universal representation}.
} single universal representation .
133. Representation Ecology
延續 TCRS-07:
E R = { N , A , S } \boxed{
\mathcal E_R
=
\{
N,A,S
\}
} E R = { N , A , S }
動態協作。
134. Representation Compiler
本文提出:
C R . \boxed{
\mathcal C_R.
} C R .
135. 它的任務不是只做 serialization
不是:
N → text → S . N
\rightarrow
\text{text}
\rightarrow
S. N → text → S .
136. 因為 text 是低成本 bridge
但常:
lossy . \boxed{
\text{lossy}.
} lossy .
137. 真正 compiler 應操作
structured state;
invariants;
typed relation;
confidence;
residual。
138. Neural-to-Structural Compiler
C N → A . \boxed{
\mathcal C_{N\rightarrow A}.
} C N → A .
139. Structural-to-Symbolic Compiler
C A → S . \boxed{
\mathcal C_{A\rightarrow S}.
} C A → S .
140. Symbolic-to-Structural Compiler
C S → A . \boxed{
\mathcal C_{S\rightarrow A}.
} C S → A .
141. Structural-to-Neural Compiler
C A → N . \boxed{
\mathcal C_{A\rightarrow N}.
} C A → N .
142. 每個 compiler 都應回傳
( output , fidelity , residual , uncertainty ) . \boxed{
(
\text{output},
\text{fidelity},
\text{residual},
\text{uncertainty}
).
} ( output , fidelity , residual , uncertainty ) .
143. 這是關鍵
translation 不應只回:
output . \boxed{
\text{output}.
} output .
144. Compiler Contract
本文提出:
input representation
output representation
preserved invariants
translation loss
unmapped residual
uncertainty
round-trip witness
145. 這可以避免 False Canonicalization
本文定義:
False Canonicalization . \boxed{
\text{False Canonicalization}.
} False Canonicalization .
146. 即:
把尚未證明等價的不同 neural states 強制映射成同一 symbolic identity。
147. 也能避免 False Neuralization
反方向:
把不同 symbolic constraint 編入 neural substrate 後,假設它們仍然完整可分。
148. Structural Layer 必須保存 distinction
否則:
compilation \boxed{
\text{compilation}
} compilation
只是:
information destruction . \boxed{
\text{information destruction}.
} information destruction .
149. Algebraic 不等於人類現成代數
這點非常重要。
本文說的:
A A A
不是:
必須全部用現成人類群論、環論、GF(2) 或某個特定代數語言。
150. 更一般:
A = structure-first representation regime . \boxed{
A
=
\text{structure-first representation regime}.
} A = structure-first representation regime .
151. 其核心問的是:
composition 怎麼定義?
equivalence 怎麼定義?
invariant 是什麼?
transformation 如何組合?
152. 所以未來 A 可能需要新數學
不一定直接是現在的:
human mathematical vocabulary . \boxed{
\text{human mathematical vocabulary}.
} human mathematical vocabulary .
153. 這也避免 Algebraic Theory Empire
否則會變成:
所有 AI 最終都是我們熟悉的代數。
154. 這又是萬法皆劍
所以:
Algebraic \boxed{
\text{Algebraic}
} Algebraic
也必須保留 boundary。
155. Structural Mathematics as Discovery
未來甚至可能:
N → A n e w \boxed{
N
\rightarrow
A_{\mathrm{new}}
} N → A new
由 neural system 自己發現新的 structural algebra。
156. 這會是一個很重要方向
因為:
A n e w A_{\mathrm{new}} A new
未必符合人類原有分類。
157. Machine-Discovered Algebra
本文提出:
Machine-Discovered Structural Algebra . \boxed{
\text{Machine-Discovered Structural Algebra}.
} Machine-Discovered Structural Algebra .
158. 它可以作為:
latent-to-canonical bridge . \boxed{
\text{latent-to-canonical bridge}.
} latent-to-canonical bridge .
159. 這時 AI 不只是學人類符號
而可能發現:
better intermediate structure . \boxed{
\text{better intermediate structure}.
} better intermediate structure .
160. 這比「LLM 接 knowledge graph」更深
因為 knowledge graph 已經預設 symbolic ontology。
161. 本文想問的是:
ontology 本身能不能從 neural 與 world interaction 中被重新編譯?
162. Ontology Compilation
本文提出:
Ontology Compilation . \boxed{
\text{Ontology Compilation}.
} Ontology Compilation .
163. 即:
N + W → A → S . \boxed{
N
+
W
\rightarrow
A
\rightarrow
S.
} N + W → A → S .
164. W 是 World Interaction
不是只靠 internal latent。
165. 因為沒有 world feedback
可能得到:
internally elegant but externally wrong structure . \boxed{
\text{internally elegant but externally wrong structure}.
} internally elegant but externally wrong structure .
166. 所以 World / Verifier 仍重要
這與 ADLM 接上。
167. Triple-Regime + World
更完整:
W ↔ N ↔ A ↔ S . \boxed{
W
\leftrightarrow
N
\leftrightarrow
A
\leftrightarrow
S.
} W ↔ N ↔ A ↔ S .
168. 甚至可形成 loop
S → A → N → W → N → A → S . \boxed{
S
\rightarrow
A
\rightarrow
N
\rightarrow
W
\rightarrow
N
\rightarrow
A
\rightarrow
S.
} S → A → N → W → N → A → S .
169. 這是一個 representation learning cycle
170. Symbolic 可以提出 hypothesis
S h . \boxed{
S_h.
} S h .
171. Algebraic 轉成 structural constraints
A h . \boxed{
A_h.
} A h .
172. Neural 實現/搜索
N h . \boxed{
N_h.
} N h .
173. World 測試
W ( N h ) . \boxed{
W(N_h).
} W ( N h ) .
174. 結果回編
W → N → A → S ′ . \boxed{
W
\rightarrow
N
\rightarrow
A
\rightarrow
S'.
} W → N → A → S ′ .
175. 這是 AI 科學循環
而不是單一 representation。
176. Representation Debt Revisited
TCRS-07 提出:
R r e p r . \boxed{
R_{\mathrm{repr}}.
} R repr .
177. Triple-Regime Architecture 的目標之一
不是完全消除 neural implicitness。
178. 而是:
selectively externalize high-value structure . \boxed{
\text{selectively externalize high-value structure}.
} selectively externalize high-value structure .
179. Selective Canonicalization
本文提出:
Selective Canonicalization . \boxed{
\text{Selective Canonicalization}.
} Selective Canonicalization .
180. 不是所有 latent state 都要 symbolify
否則:
C c a n o n i c a l → ∞ . C_{\mathrm{canonical}}\rightarrow\infty. C canonical → ∞.
181. 應只對:
high reuse;
high risk;
high causal centrality;
high verification value;
的 structure canonicalize。
182. Canonicalization Priority
定義:
P C ( x ) = F ( reuse , risk , causal centrality , verification value ) . \boxed{
P_C(x)
=
F(
\text{reuse},
\text{risk},
\text{causal centrality},
\text{verification value}
).
} P C ( x ) = F ( reuse , risk , causal centrality , verification value ) .
183. 這讓 Symbolic 不會吞掉 Neural
也避免回到 GOFAI 的顯式建構天價成本。
184. 這是一個重要平衡
implicit where cheap + explicit where valuable . \boxed{
\text{implicit where cheap}
+
\text{explicit where valuable}.
} implicit where cheap + explicit where valuable .
185. Structural layer 負責判斷哪裡值得顯式化
這可能是:
A A A
最重要角色之一。
186. Adaptive Explicitness
本文提出:
Adaptive Explicitness . \boxed{
\text{Adaptive Explicitness}.
} Adaptive Explicitness .
187. 即:
explicitness 不是 binary,而是依 task 動態分配。
188. 某些 knowledge
可以留:
N . \boxed{
N.
} N .
189. 某些 structure
升到:
A . \boxed{
A.
} A .
190. 某些 high-value knowledge
進:
S . \boxed{
S.
} S .
191. Representation Promotion
本文提出:
N → A → S \boxed{
N
\rightarrow
A
\rightarrow
S
} N → A → S
作為 promotion path。
192. Representation Demotion
也可以:
S → A → N . \boxed{
S
\rightarrow
A
\rightarrow
N.
} S → A → N .
193. 為什麼要 demote?
因為某些 symbolic structure:
可能更適合回到 learned regime。
194. 所以 representation regime 是動態的
不是知識一旦 symbolized 就永遠 symbolized。
195. Dynamic Representation State
每個 knowledge object:
k i k_i k i
可以有:
r i ( t ) ∈ { N , A , S } . \boxed{
r_i(t)\in\{N,A,S\}.
} r i ( t ) ∈ { N , A , S } .
196. 甚至是混合態
r i ( t ) = ( w N , w A , w S ) . \boxed{
r_i(t)
=
(w_N,w_A,w_S).
} r i ( t ) = ( w N , w A , w S ) .
197. 這與 TCRS-07 的 Representation Ecology 完全接軌
198. 理論完備性再看一次
Symbolic 看起來能:
enumerate and verify \boxed{
\text{enumerate and verify}
} enumerate and verify
所以有類完備感。
199. Neural 看起來能:
approximate arbitrary patterns \boxed{
\text{approximate arbitrary patterns}
} approximate arbitrary patterns
所以也有類完備感。
200. Algebraic 看起來能:
reduce systems to structural relations \boxed{
\text{reduce systems to structural relations}
} reduce systems to structural relations
所以也有類完備感。
201. 但三者的「完備」不是同一種
本文提出:
Completeness Type . \boxed{
\text{Completeness Type}.
} Completeness Type .
202. Symbolic Completeness
偏:
explicit derivational completeness . \boxed{
\text{explicit derivational completeness}.
} explicit derivational completeness .
203. Neural Completeness
偏:
approximation / representation coverage . \boxed{
\text{approximation / representation coverage}.
} approximation / representation coverage .
204. Algebraic Completeness
偏:
structural closure under transformations . \boxed{
\text{structural closure under transformations}.
} structural closure under transformations .
205. 不同 completeness 不能互相偷換
C S ≠ C N ≠ C A . \boxed{
C_S
\neq
C_N
\neq
C_A.
} C S = C N = C A .
206. 這也是學派歧視的來源之一
每派都用自己的 completeness metric 看別人。
207. Symbolic 會覺得 Neural 不完備
因為:
無法明確 trace。
208. Neural 會覺得 Symbolic 不完備
因為:
無法涵蓋 open-ended messy world。
209. Algebraic 可能覺得兩者都太 representation-bound
因為:
沒有抽出 invariant。
210. 三方都可能是對的
只是在不同 metric。
211. Completeness-Metric Mismatch
本文提出:
CMM = Completeness-Metric Mismatch . \boxed{
\text{CMM}
=
\text{Completeness-Metric Mismatch}.
} CMM = Completeness-Metric Mismatch .
212. 當不同 paradigm 用不同 completeness definition
卻假裝在比較同一個:
complete AI , \boxed{
\text{complete AI},
} complete AI ,
就產生假衝突。
213. Neutral Completeness Space
未來比較應建立:
C ⃗ = ( C d e r i v , C a p p r o x , C s t r u c t , C w o r l d ) . \boxed{
\vec C
=
(
C_{\mathrm{deriv}},
C_{\mathrm{approx}},
C_{\mathrm{struct}},
C_{\mathrm{world}}
).
} C = ( C deriv , C approx , C struct , C world ) .
214. World Completeness
還要加入:
C w o r l d \boxed{
C_{\mathrm{world}}
} C world
也就是:
系統和實際世界的匹配。
215. 因為內部完備不等於外部正確
Internal Completeness ≠ World Adequacy . \boxed{
\text{Internal Completeness}
\neq
\text{World Adequacy}.
} Internal Completeness = World Adequacy .
216. Triple-Regime Evaluation
因此未來 architecture 應測:
learning coverage;
structural conservation;
symbolic addressability;
world adequacy;
repair cost;
verification cost。
217. 這比 benchmark accuracy 高解析很多
218. TCRS-08 的核心 Green Model
不是:
S wins \boxed{
S\text{ wins}
} S wins
也不是:
N wins . N\text{ wins}. N wins .
219. 更不是:
A wins . A\text{ wins}. A wins .
220. 而是:
regime selection becomes task- and bottleneck-dependent . \boxed{
\text{regime selection becomes task- and bottleneck-dependent}.
} regime selection becomes task- and bottleneck-dependent .
221. Future AI Architecture
可能更接近:
A ∗ = R ( Q , B , W , R ⃗ P ) \boxed{
\mathcal A^\ast
=
\mathcal R(
Q,
B,
W,
\vec R_P
)
} A ∗ = R ( Q , B , W , R P )
動態選擇 representation regime。
222. Q 是 task
Q . Q. Q .
223. B 是 bottleneck
B . B. B .
224. W 是 world constraints
W . W. W .
225. Debt vector
R ⃗ P . \vec R_P. R P .
226. Router 決定
( w N , w A , w S ) . \boxed{
(w_N,w_A,w_S).
} ( w N , w A , w S ) .
227. 這就是 Representation Router
本文提出:
R r e p . \boxed{
\mathcal R_{\mathrm{rep}}.
} R rep .
228. Representation Router 的學習目標
不是:
永遠偏某一派。
而是:
min ( cost + error + debt ) \boxed{
\min
(
\text{cost}
+
\text{error}
+
\text{debt}
)
} min ( cost + error + debt )
在保持 capability 下。
229. 形式:
w ⃗ ∗ = arg min w ⃗ [ C ( w ⃗ ) + E ( w ⃗ ) + R ( w ⃗ ) ] . \boxed{
\vec w^\ast
=
\arg\min_{\vec w}
\left[
C(\vec w)
+
E(\vec w)
+
R(\vec w)
\right].
} w ∗ = arg w min [ C ( w ) + E ( w ) + R ( w ) ] .
230. 這就完全離開「學派戰爭」
問題變成 engineering / epistemic optimization。
231. 與 TCRS-07 的 Bottleneck-Neutral Comparison 接軌
不是問:
Neural 怎麼解?
而是:
哪種 representation 組合最適合這個問題? \boxed{
\text{哪種 representation 組合最適合這個問題?}
} 哪種 representation 組合最適合這個問題?
232. 本文第一原則:Representation-Regime Principle
Symbolic, neural, and algebraic systems should first be treated as representation regimes, not exclusive ontologies of intelligence . \boxed{
\text{Symbolic, neural, and algebraic systems should first be treated as representation regimes, not exclusive ontologies of intelligence}.
} Symbolic, neural, and algebraic systems should first be treated as representation regimes, not exclusive ontologies of intelligence .
233. 第二原則:Identity–Geometry–Invariant Principle
Identity, geometry, and invariant structure are distinct computational assets that need not reside in one representation . \boxed{
\text{Identity, geometry, and invariant structure are distinct computational assets that need not reside in one representation}.
} Identity, geometry, and invariant structure are distinct computational assets that need not reside in one representation .
234. 第三原則:Structural-Mediation Principle
A structural layer may mediate neural–symbolic translation more faithfully than direct serialization . \boxed{
\text{A structural layer may mediate neural–symbolic translation more faithfully than direct serialization}.
} A structural layer may mediate neural–symbolic translation more faithfully than direct serialization .
235. 第四原則:Bidirectional Compilation Principle
N ↔ A ↔ S \boxed{
N
\leftrightarrow
A
\leftrightarrow
S
} N ↔ A ↔ S
應支援雙向而非單向投影。
236. 第五原則:Residual Preservation Principle
No representation conversion should claim completeness without tracking residual information . \boxed{
\text{No representation conversion should claim completeness without tracking residual information}.
} No representation conversion should claim completeness without tracking residual information .
237. 第六原則:Structural Conservation Principle
Cross-regime translation should preserve selected invariants rather than literal representation identity . \boxed{
\text{Cross-regime translation should preserve selected invariants rather than literal representation identity}.
} Cross-regime translation should preserve selected invariants rather than literal representation identity .
238. 第七原則:Adaptive Explicitness Principle
Not all learned structure should be made explicit; explicitness should be allocated where its value exceeds its cost . \boxed{
\text{Not all learned structure should be made explicit; explicitness should be allocated where its value exceeds its cost}.
} Not all learned structure should be made explicit; explicitness should be allocated where its value exceeds its cost .
239. 第八原則:Completeness-Type Principle
Different representation regimes possess different kinds of completeness and must not be compared using one regime’s metric alone . \boxed{
\text{Different representation regimes possess different kinds of completeness and must not be compared using one regime's metric alone}.
} Different representation regimes possess different kinds of completeness and must not be compared using one regime’s metric alone .
240. 第九原則:World-Adequacy Principle
Internal structural elegance across representations does not replace external world validation . \boxed{
\text{Internal structural elegance across representations does not replace external world validation}.
} Internal structural elegance across representations does not replace external world validation .
241. 第十原則:Representation-Ecology Principle
Future AI may be better modeled as a dynamic ecology of representation regimes than as a single universal representation . \boxed{
\text{Future AI may be better modeled as a dynamic ecology of representation regimes than as a single universal representation}.
} Future AI may be better modeled as a dynamic ecology of representation regimes than as a single universal representation .
242. 本文不主張事項
本文不主張:
Symbolic、Neural、Algebraic 是完整且互斥的 AI 分類;
Algebraic layer 一定是 neuro-symbolic 唯一缺失層;
所有 neural state 都能被 canonicalize;
所有 symbolic structure 都能無損編入 neural substrate;
現有代數語言足以描述未來 AI;
Neural 應被 Symbolic 取代;
Symbolic 應被 Algebraic 取代;
Triple-Regime Architecture 一定優於 pure neural;
representation-independent invariant 一定存在於所有任務;
AI 最終必須使用人類可讀 symbolic language。
本文只主張:
the competition between symbolic, neural, and algebraic approaches may be better reframed as a problem of cross-regime structure preservation, compilation, and task-dependent routing . \boxed{
\text{the competition between symbolic, neural, and algebraic approaches may be better reframed as a problem of cross-regime structure preservation, compilation, and task-dependent routing}.
} the competition between symbolic, neural, and algebraic approaches may be better reframed as a problem of cross-regime structure preservation, compilation, and task-dependent routing .
243. 與 TCRS-09 的接口
下一篇將完成主系列 closure:
《從學派到共同底空間:多解析度理論比較與跨範式編譯的統一方法》
它會把:
一劍萬法/萬法一劍;
多層理論域;
理解解析度;
對話假對齊;
Shared-Substrate Occlusion;
Theory Empire;
Bottleneck Migration;
Symbolic–Neural–Algebraic;
統一成一個跨學科方法論。
244. 結論
Symbolic、Neural 與 Algebraic 之間真正的差異,很大一部分可能不是:
哪一個才是真正智能。 \boxed{
\text{哪一個才是真正智能。}
} 哪一個才是真正智能。
而是:
它們選擇了不同的 computational coordinates。 \boxed{
\text{它們選擇了不同的 computational coordinates。}
} 它們選擇了不同的 computational coordinates 。
Symbolic 重視:
Identity + Addressability + Explicit Rule . \boxed{
\text{Identity}
+
\text{Addressability}
+
\text{Explicit Rule}.
} Identity + Addressability + Explicit Rule .
Neural 重視:
Distributed Geometry + Adaptation + Approximation . \boxed{
\text{Distributed Geometry}
+
\text{Adaptation}
+
\text{Approximation}.
} Distributed Geometry + Adaptation + Approximation .
Algebraic / Structural 重視:
Invariant + Composition + Transformation + Equivalence . \boxed{
\text{Invariant}
+
\text{Composition}
+
\text{Transformation}
+
\text{Equivalence}.
} Invariant + Composition + Transformation + Equivalence .
因此真正困難的問題不是:
哪一派應該消滅另外兩派?
而是:
同一結構在不同表示制度間轉換時,哪些東西應該被保留? \boxed{
\text{同一結構在不同表示制度間轉換時,哪些東西應該被保留?}
} 同一結構在不同表示制度間轉換時,哪些東西應該被保留?
如果只有:
N → S , N\rightarrow S, N → S ,
很容易把 neural geometry 與 uncertainty 強行壓成 symbolic object。
如果只有:
S → N , S\rightarrow N, S → N ,
又可能把 explicit identity 與 constraint 編進一個難以回收的 distributed state。
所以本文提出:
N ↔ A ↔ S . \boxed{
N
\leftrightarrow
A
\leftrightarrow
S.
} N ↔ A ↔ S .
讓:
A A A
不是第三個學派王座,
而是:
structure-preserving mediation layer . \boxed{
\text{structure-preserving mediation layer}.
} structure-preserving mediation layer .
它負責問:
什麼是同一個 structure?
哪些 transformation 可以組合?
哪些 representation 應視為等價?
哪些 invariant 在 coordinate change 後仍然存在?
哪些 residual 不能被任何一方吸收?
真正好的未來 AI 不必把所有東西都 symbolify。
也不必把所有東西永遠留在 latent space。
它可以:
implicit where cheap + structural where compositional + explicit where valuable . \boxed{
\text{implicit where cheap}
+
\text{structural where compositional}
+
\text{explicit where valuable}.
} implicit where cheap + structural where compositional + explicit where valuable .
也就是:
Adaptive Explicitness . \boxed{
\text{Adaptive Explicitness}.
} Adaptive Explicitness .
這樣一來,Symbolic、Neural、Algebraic 不再是三個互斥世界。
它們更像:
同一計算結構的三種座標系。 \boxed{
\text{同一計算結構的三種座標系。}
} 同一計算結構的三種座標系。
而真正值得研究的,便從:
Which school wins? \boxed{
\text{Which school wins?}
} Which school wins?
變成:
What survives when we change representation? \boxed{
\text{What survives when we change representation?}
} What survives when we change representation?
這就是 TCRS-08 的核心。
版本紀錄
v0.1.0 — 2026-10-06
建立 TCRS 系列 Paper 08;
將 Symbolic / Neural / Algebraic 重新定義為三種 representation regimes;
建立 Symbolic 的 Identity / Addressability / Explicit Rule / Verification 優勢;
建立 Neural 的 Distributed Representation / Adaptation / Approximation / High-Dimensional Association 優勢;
建立 Algebraic / Structural 的 Invariant / Composition / Transformation / Equivalence 優勢;
提出 Identity–Geometry–Invariant Triangle;
正式提出 Structural Mediation Hypothesis;
建立 Neural → Algebraic → Symbolic 與反向路徑;
正式提出 Bidirectional Representation Compilation;
提出 Representation-Preserving Translation;
建立 Canonicalization Witness 與 False Canonicalization;
提出 Structural Conservation;
將 representation 視為 computational coordinate systems;
建立 Representation-Independent Invariant;
提出 Regime-Relative Completeness;
建立 Mutual Regime Absorption;
定義 Shared Structural Substrate;
提出 Representation Round-Trip 與 Round-Trip Error;
建立 Addressability Gap、Geometry Gap 與 Structural Gap;
正式提出 Triple-Regime Architecture;
提出 Dynamic Regime Routing;
建立 Representation Compiler Contract;
將 Algebraic 抽象為 structure-first regime,而非限定現成人類代數;
提出 Machine-Discovered Structural Algebra;
建立 Ontology Compilation;
將 Triple-Regime 與 World / Verifier 接軌;
提出 Selective Canonicalization、Adaptive Explicitness、Representation Promotion / Demotion;
建立 Completeness-Type 與 Completeness-Metric Mismatch;
提出 Representation Router;
建立十項核心原則;
為 TCRS-09 統一 closure 建立最後理論前置。