實驗EXP-2026-0017v0.1
PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性
在 v0.9 圖冊中 A、B、C 三個 chart 都有效的凍結區域上,直接轉換 A → C 與組合 A → B → C 在潛在 JS 上一致到 ≈ 10⁻⁵(shuffled-pair 控制 ≈ 10⁻¹),且兩條路徑各自對 chart C 都準確(潛在 JS ≈ 0.0045–0.0054)。四個家族在主要 seed 與四個次要 seed 下都通過。N2 只有一個三重重疊幾何(q = 0.24),見證較窄。
假設
hypothesis- T_AC ≈ T_BC ∘ T_AB on held-out triple-overlap data.
設定
model_idsdataset_idsbenchmark_idssoftware_environment- Python; deterministic seeded generators; no network, no LLM.
執行
run_count- 5
random_seeds- 20260909
- 3
- 5
- 7
- 11
controls- shuffled-target mapping
- constant prediction
- broken composition (v0.6+)
- target-local refit (v0.7+)
metricsverdict- ROBUST_ORIENTED_COCYCLE_COHERENCE_ON_FROZEN_TRIPLE_OVERLAP
N0overlap_q- 0.24
- 0.3
direct_vs_composed_latent_js- 3.5e-05
direct_to_C- 0.00446
composed_to_C- 0.004627
real_over_shuffled- 0.000346
N3overlap_q- 0.2
- 0.24
- 0.3
direct_vs_composed_latent_js- 2.65e-05
secondary_pass_rate- 1.0
詮釋
interpretation- Pairwise transition quality may be uneven globally while cocycle composition is highly coherent locally on triple overlap — consistent with coordinate compatibility being a local overlap property. Too early for a coordinate groupoid or manifold.
限制
limitations- Evidence volume differs by family; N2's triple overlap is a single geometry.
重現
reproduction_instructions- Extract the version's FINAL bundle; python -m pytest -q; run the version's primary script with the recorded seed; docs/PACC_LAB_v0.N_RESULTS.md and docs/EXPERIMENT_PROTOCOL_v0.N.md are inside the bundle.
結果
| 來源 | 關係 | 目標 | 狀態 | ID |
|---|---|---|---|---|
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | produces | RST-2026-0011 v0.10 cocycle:路徑分歧 ≈ 10⁻⁵ vs 控制 ≈ 10⁻¹ | ACTIVE | REL-2026-0276 |
記錄欄位
completed_at- 2026-09-09
關係
| 來源 | 關係 | 目標 | 狀態 | ID |
|---|---|---|---|---|
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | runs_on | SYS-2026-0002 PACC-Lab——收斂猜想的微型實驗室 | ACTIVE | REL-2026-0265 |
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | uses_benchmark | BEN-2026-0001 PACC 微型實驗室協定 v0.1(凍結門檻) | ACTIVE | REL-2026-0266 |
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | uses_dataset | DAT-2026-0001 PACC 合成證據世界 | ACTIVE | REL-2026-0267 |
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | uses_model | MOD-2026-0001 N0——帶號支持 | ACTIVE | REL-2026-0268 |
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | uses_model | MOD-2026-0002 N1——序數錦標賽 | ACTIVE | REL-2026-0269 |
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | uses_model | MOD-2026-0003 N2——帶號圖 | ACTIVE | REL-2026-0270 |
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | uses_model | MOD-2026-0004 N3——約束競爭 | ACTIVE | REL-2026-0271 |
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | uses_model | MOD-2026-0005 Bayesian 參考(精確後驗/Beta-Bernoulli/聯合共同因/HMM) | ACTIVE | REL-2026-0272 |
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | tests | THY-2026-0005 PACC 猜想——四層收斂階梯 | ACTIVE | REL-2026-0273 |
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | extends | EXP-2026-0016 PACC-Lab v0.9——三錨點概率座標圖冊 | ACTIVE | REL-2026-0274 |
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | produced | ART-2026-0030 PACC-Lab v0.10 Cocycle Coherence FINAL artifact://evemisslab/adaptive-epistemic-systems/PACC-Lab_v0.10_Cocycle_Coherence_FINAL.zip | ACTIVE | REL-2026-0275 |
EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | produces | RST-2026-0011 v0.10 cocycle:路徑分歧 ≈ 10⁻⁵ vs 控制 ≈ 10⁻¹ | ACTIVE | REL-2026-0276 |
EXP-2026-0018 PACC-Lab v0.11——雙向 cocycle 與反向一致性 | extends | EXP-2026-0017 PACC-Lab v0.10——凍結三重重疊上的定向 cocycle 一致性 | ACTIVE | REL-2026-0287 |
歷史與來源歷程
- Canonical URL
- https://evemisslab.com/ai/experiments/EXP-2026-0017/
- 快照
AI-SNAPSHOT-v0.1-fe85b9694a45- 來源歷程
source- EveMissLab research collection: Adaptive Epistemic Systems (真本體論13)
extracted_by- Splice (Claude Code), reading the canonical UTF-8 sources and each lab's own result reports
extracted_at- 2026-09-11
generator- tools/extract_aes/extract.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