實驗EXP-2026-0015v0.1
PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域
在 q* = 0.20 擬合的一個組合座標,原封不動地套用到九點掃描 q ∈ {0.06 … 0.42}(叢集內正確性相關從 ≈ 0.203 升到 ≈ 0.502)。N0/N1/N3 通過到 q = 0.33、從 0.36 起失敗;N2 通過到 0.30、從 0.33 起失敗。在邊緣先失去的是負控制判別餘裕,而絕對潛在誤差仍小;目標本地重擬合處處 9/9 通過。四個次要 seed 對 N0/N1/N3 精確重現吸引域,N2 在 0.33 的邊緣對 seed 敏感。
假設
hypothesis- A frozen probability-like coordinate has a finite transfer basin whose width depends on the family.
設定
model_idsdataset_idsbenchmark_idssoftware_environment- Python; deterministic seeded generators; no network, no LLM.
執行
run_count- 2
random_seeds- 20260909
- 4 secondary seeds
controls- shuffled-target mapping
- constant prediction
- broken composition (v0.6+)
- target-local refit (v0.7+)
metricsverdict- FAMILY_SPECIFIC_TRANSFER_BASINS
basins_gridN0_N1_N3- [0.06, 0.33]
N2- [0.06, 0.30]
pass_fractionN0- 7/9
N2- 6/9
N3- 7/9
local_refit- 9/9
edge_N0_q0.36task_js- 0.0488
latent_js- 0.0274
real_over_control- 0.858
fails- negative-control discrimination
詮釋
interpretation- An atlas/basin picture: large regions share one low-complexity coordinate; different internal dynamics meet transfer boundaries at different places. High-q failure is not evidence that probability-like coordinates cease to exist there.
限制
limitations- Grid points only, not continuous coverage; no single global coordinate over all dependence strengths.
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-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | produces | RST-2026-0010 v0.8 吸引域圖 | ACTIVE | REL-2026-0252 |
記錄欄位
completed_at- 2026-09-09
關係
| 來源 | 關係 | 目標 | 狀態 | ID |
|---|---|---|---|---|
EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | runs_on | SYS-2026-0002 PACC-Lab——收斂猜想的微型實驗室 | ACTIVE | REL-2026-0241 |
EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | uses_benchmark | BEN-2026-0001 PACC 微型實驗室協定 v0.1(凍結門檻) | ACTIVE | REL-2026-0242 |
EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | uses_dataset | DAT-2026-0001 PACC 合成證據世界 | ACTIVE | REL-2026-0243 |
EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | uses_model | MOD-2026-0001 N0——帶號支持 | ACTIVE | REL-2026-0244 |
EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | uses_model | MOD-2026-0002 N1——序數錦標賽 | ACTIVE | REL-2026-0245 |
EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | uses_model | MOD-2026-0003 N2——帶號圖 | ACTIVE | REL-2026-0246 |
EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | uses_model | MOD-2026-0004 N3——約束競爭 | ACTIVE | REL-2026-0247 |
EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | uses_model | MOD-2026-0005 Bayesian 參考(精確後驗/Beta-Bernoulli/聯合共同因/HMM) | ACTIVE | REL-2026-0248 |
EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | tests | THY-2026-0005 PACC 猜想——四層收斂階梯 | ACTIVE | REL-2026-0249 |
EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | extends | EXP-2026-0014 PACC-Lab v0.7——不重新擬合的跨幾何座標轉移 | ACTIVE | REL-2026-0250 |
EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | produced | ART-2026-0028 PACC-Lab v0.8 Transfer Basin FINAL artifact://evemisslab/adaptive-epistemic-systems/PACC-Lab_v0.8_Transfer_Basin_FINAL.zip | ACTIVE | REL-2026-0251 |
EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | produces | RST-2026-0010 v0.8 吸引域圖 | ACTIVE | REL-2026-0252 |
EXP-2026-0016 PACC-Lab v0.9——三錨點概率座標圖冊 | extends | EXP-2026-0015 PACC-Lab v0.8——相關度掃描下的凍結座標轉移吸引域 | ACTIVE | REL-2026-0263 |
歷史與來源歷程
- Canonical URL
- https://evemisslab.com/ai/experiments/EXP-2026-0015/
- 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