理論THY-2026-0101v0.1
回合分解與外部無迴圈智能
把「一輪」分解成向量 T =（U 使用者回合、G 生成軌跡、I 模型呼叫、L 外部回饋迴圈、R 重試、S 選擇／驗證、P 物理執行軌跡）。最乾淨的 single pass 是 U=1、G=1、R=1、L=0、S=0，排除的只有「行動→新證據→重規劃」，不排除自回歸的序列計算。LOOP 分五類（工具、環境、驗證器、重試、候選／選擇），能力分四層（intrinsic、elicited、system、product），任何「一輪完成」的宣稱都要投影到事件向量 (Q, U, G, I, L, R, S, T, E, V_CST)。
定義
definitions- Turn Decomposition Framework T = (U, G, I, L, R, S, P).
- Externally Loopless Intelligence (ELI): during a task the system cannot re-aim on external world, tool output, an independent verifier, another candidate or a retry.
- Interaction compression C_U = internal interactions / U.
- Loop taxonomy L = (L_T, L_E, L_V, L_R, L_C).
前提
assumptions- Autoregressive generation is internal sequential computation, not an external loop.
- Selection over candidates is itself a capability source.
主張
claims- UserTurn ≠ PhysicalTurn; UserTurn ≠ ModelInvocation; ModelInvocation ≠ GenerationTrajectory.
- NoExternalLoop ≠ NoSequentialComputation.
- Pass@k ≠ Pass@1; BestOfN ≠ MedianExperience.
- InteractionCompression ≠ ComputationCompression; SameFinalQuality ≠ SameIntelligenceEfficiency.
形式化
formalization- Single-pass condition: U=1, G=1, R=1, L=0, S=0 → Q_SP.
- P(at least one success in k rollouts) = 1 − (1 − p)^k.
- Unified event vector 𝔈 = (Q, U, G, I, L, R, S, T, E, V_CST).
預測
predictions- Systems reported as 'one turn' will separate by orders of magnitude on (L, R, S, V_CST, E) once the event vector is recorded.
證偽／失敗條件
falsification_conditions- If operational accounting of U, G, I, L, R, S adds nothing to predicting cost or quality beyond the visible answer, the decomposition is idle.
已知限制
known_limitations- Twelve invariants and a vocabulary; no measurement of its own — the vector is filled by Papers 02–09.
關係
| 來源 | 關係 | 目標 | 狀態 | ID |
|---|---|---|---|---|
RES-2026-0101 執行與物理線——一個答案在回合、語意工作、能量與計算時空上的代價 | develops | THY-2026-0101 回合分解與外部無迴圈智能 | ACTIVE | REL-2026-0358 |
THY-2026-0102 μI——最小智能語意執行單位（候選理論） | extends | THY-2026-0101 回合分解與外部無迴圈智能 | ACTIVE | REL-2026-0368 |
THY-2026-0109 鷹架能力紀錄：SSR、SDR、SCM 與消融階梯 | extends | THY-2026-0101 回合分解與外部無迴圈智能 | ACTIVE | REL-2026-0374 |
PAP-2026-0101 一輪到底是一輪什麼？：使用者回合、隱藏 LOOP 與單次智能的重新定義 | formalizes | THY-2026-0101 回合分解與外部無迴圈智能 | ACTIVE | REL-2026-0392 |
歷史與來源歷程
- Canonical URL
- https://evemisslab.com/ai/theory/THY-2026-0101/
- 快照
AI-SNAPSHOT-v0.1-fe85b9694a45- 來源歷程
source- EveMissLab research collection: Intelligence Physical Metrology (真本體論13)
extracted_by- Splice (Claude Code), reading the canonical UTF-8 sources and each package's own reports
extracted_at- 2026-09-11
generator- tools/extract_all.py
claim_boundary- status, evidence level and result type follow the source artifact's own stated claim boundary; nothing is upgraded beyond what the report supports