55-Decision & Change Log Template v1.0 | Chapter 5 Option Gathering & Comparison

Home ／ Docs-Technical WhitePaper ／ 55-Decision & Change Log Template v1.0

Chapter 5 Option Gathering & Comparison

I. Chapter Goals & Boundaries (Mandatory)

• Goal: build a set of comparable and adjudicable options, enforce unified evidence caliber, scoring, and gate binding, and output conclusions consistent across human × machine views.
• Boundary: covers technical/caliber/process/data-contract/implementation-binding options only; organization, HR, and budget approval are out of scope.

II. Option Sourcing Framework (Mandatory)

• Sources: internal proposals, external literature, existing implementations, baseline.
• Minimum count: ≥2 comparable candidates (Option A/B/...). If only one exists, record evidence of non-substitutability here.
• Baseline: define Option Base (current caliber/implementation) as the comparator.
• Naming: use the “object—attribute—caliber” triplet, e.g., n_eff.medium=intergalactic, c_ref.band=L2, T_fil.regime=linear.

III. Evidence & Uncertainty List (Mandatory)

• Evidence elements: dataset versions, experiment/compute scripts, statistical methods, confidence intervals, and a replay path (script@commit).
• Dimensional caliber: all formulas and parameters must pass check_dim; specify units and dimensions at first occurrence.
• Reproducibility: provide minimal replay commands and environment notes (container/deps) and register them in the “Verification” section.

IV. Evaluation Dimensions & Scoring Rules (Mandatory)

• Dimensions (default set): performance, cost, time, risk, dependency, reproducibility, compatibility-with-prod.
• Quantification: map each raw metric to [0,1] as s_i; weights w_i≥0, with ∑ w_i = 1.
• Total score: S = ∑_i ( w_i · s_i ).
• Uncertainty discount: if the estimated std. dev. is σ_i, use s_i' = clamp( s_i - k·σ_i, 0, 1 ), default k=1.
• Constraint: hard gates take precedence over scoring; options failing hard gates are not ranked.

V. Comparable Narrative Template (Mandatory)

• Problem & Goals (1–3 paragraphs): what is optimized/constrained, measurable targets.
• Assumptions & Boundaries (bulleted): key assumptions, applicability, exclusions.
• Method Summary (1–3 paragraphs): core idea and differences vs baseline.
• Evidence Summary (table + citations): key experiments/data and significance.
• Risks & Rollback (bulleted): failure modes, triggers, rollback path.
• Baseline/Peer Comparison (table + prose): dimension scores and pros/cons conclusion.

VI. Option Comparison Matrix (copy-ready)

VII. Criteria & Gate Binding (Mandatory)

• Naming: gate_<metric><comparator><threshold>@<window>, e.g., gate_accuracy>=0.99@7d, gate_latency<=2h@7d.
• Policy: hard (failure eliminates option) and soft (may proceed with Approver signature).
• Output: in the DR “Implementation & Verification” section, provide gate data sources, stats methods, and script locators.

VIII. Path/Formula Consistency (Mandatory)

1. If arrival-time criteria are involved, use the unified forms:
   • Constant factored: T_arr = ( 1 / c_ref ) * ( ∫ n_eff d ell )
   • General form: T_arr = ( ∫ ( n_eff / c_ref ) d ell )
2. When T_arr appears, in the same or adjacent paragraph explicitly declare path gamma(ell) and measure d ell; all dimensional checks pass via check_dim.
3. No mixing: T_fil ≠ T_trans, n ≠ n_eff, c ≠ c_ref; no Chinese in formulas/symbols/definitions.

IX. Machine-Readable Schema (YAML; JSON equivalent, copy-ready)

options:

base:

name: "Base"

assumptions: []

evidence:

datasets: []

scripts: []

metrics: { gate_accuracy: "0.97@7d", gate_latency: "2.5h@7d" }

candidates:

- name: "Option A"

assumptions: []

evidence:

datasets: ["CMB-Set@v3"]

scripts: ["arrive_time_check.py@a1b2c3"]

stats: { method: "bootstrap", ci: "95%" }

gates:

hard: ["gate_accuracy>=0.99@7d","gate_latency<=2h@7d"]

soft: ["compat_rate>=0.995@replay"]

scores:

normalized:

performance: 0.992

cost: 0.90

time: 0.72

risk: 0.92

dependency: 0.80

reproducibility: 0.99

compatibility: 0.995

sigma:

performance: 0.004

formula:

total: "S = sum_i(w_i * max(0, min(1, s_i - k*sigma_i)))"

k: 1.0

weights:

performance: 0.30

cost: 0.15

time: 0.10

risk: 0.15

dependency: 0.10

reproducibility: 0.10

compatibility: 0.10

decision_rules:

hard_gates: ["gate_accuracy>=0.99@7d","gate_latency<=2h@7d"]

tie_breakers: ["risk","compatibility","cost"]

X. Minimal Filled Example (copy-ready)

options:

base: { name: "Base", evidence: { datasets:["cmb_v2"], scripts:["baseline.py@9f8e7d"] } }

candidates:

- name: "A - General T_arr"

assumptions: ["Allow spatially varying c_ref and n_eff"]

evidence:

datasets: ["cmb_v3","lens_v1"]

scripts: ["arrive_time_check.py@a1b2c3"]

stats: { method:"bootstrap", ci:"95%" }

gates:

hard: ["gate_accuracy>=0.99@7d","gate_latency<=2h@7d"]

scores:

normalized: { performance:0.995, cost:0.90, time:0.80, risk:0.93, dependency:0.80, reproducibility:0.99, compatibility:0.995 }

sigma: { performance:0.003 }

- name: "B - Constant-Factored T_arr"

assumptions: ["c_ref approx. constant within segment"]

evidence:

datasets: ["cmb_v3"]

scripts: ["arrive_time_check.py@a1b2c3"]

gates:

hard: ["gate_accuracy>=0.99@7d","gate_latency<=2h@7d"]

scores:

normalized: { performance:0.992, cost:1.10, time:0.85, risk:0.90, dependency:0.70, reproducibility:0.98, compatibility:0.992 }

weights: { performance:0.30, cost:0.15, time:0.10, risk:0.15, dependency:0.10, reproducibility:0.10, compatibility:0.10 }

summary:

ranking: ["A","B"]

rationale:

- "A leads in performance, reproducibility, and compatibility; passes all hard gates; cost below baseline."

- "B is close in performance but higher cost and wider dependency surface."

recommendation: "Select A; record exceptions that require constant factoring as explicit clauses."

XI. Cross-References & Citation Style (Mandatory)

1. Fixed in-text format: “See 《 vX.Y》 Ch.x S/P/M/I…”, anchors preferred over whole volumes.
2. Provide a machine-readable list in references.see within the DR (examples):
   • "EFT.WP.Core.Terms v1.0:P10-3"
   • "EFT.WP.Core.Equations v1.1:S20-1"
   • "EFT.WP.Core.Metrology v1.0:check_dim"
   • "EFT.WP.Core.DataSpec v1.0:I30-2"