fpf.tools
Library / First Principles Framework (FPF) - Core Conceptual Specification
Jump to passage
In this reading
Source · lines 118771–118816

Link to current text

Published source confirmed at last check

Source changed 2026-10-03 02:22:15 UTC · snapshot created 2026-10-03 03:38:22 UTC · last check 2026-10-03 04:25:17 UTC

G.11:11 - SoTA-Echoing — Post‑2015 practices aligned (informative)

Each entry follows: claim → practice → source → alignment → adoption status.

0. QD currentness requires visible survey support. Practice: current QD work is surveyed as approaches, applications, and challenges, with archive, diversity, descriptor, and evaluator-currentness concerns still live. Source: A survey on Quality-Diversity optimization: Approaches, applications, and challenges, Swarm and Evolutionary Computation 2026, DOI 10.1016/j.swevo.2025.102240. Alignment: RefreshCurrentnessLine@Context may name selected set, Front, Q-front, ExplorationArchive, Archive, portfolio lineage, descriptor or distance edition, and path-slice scope, while C.18, C.19, and G.5 keep archive, pool, and selected-set meanings. Adoption: Adopt and bound (survey support changes refresh currentness fields and boundaries; it is not the governing ontology source).

0a. Open-ended engineering outputs need source and evaluator currentness. Practice: self-improving-agent, AlphaEvolve-style, and DeepEvolve-style lines use generated variants, external knowledge, evaluators, tests, archives, and empirical validation. Source: Darwin Godel Machine arXiv:2505.22954, AlphaEvolve arXiv:2506.13131, and DeepEvolve-style deep-research augmentation arXiv:2510.06056. Alignment: G.11 refresh records carry source, evaluator, descriptor, policy, edition, lineage, and report refs; generated method text, evaluator success, and archive update keep their subject patterns. Adoption: Adopt and adapt (refresh tracks currentness and smallest affected scope; it does not accept generated text as proof, gate passage, or performed work).

  1. Continuous refresh is necessary in deployed evaluation pipelines. Practice: production ML systems use monitoring, retraining, and reevaluation triggers and insist on reproducibility hooks. Source: Breck et al., The ML Test Score (arXiv:1706.04599, 2017); Amershi et al., Software Engineering for Machine Learning (ICSE-SEIP 2019). Alignment: G.11 formalizes triggers as typed causes and forces edition and policy pins for replay. Adoption: Adopt and adapt (adapted to id-based, PathSlice-scoped refresh rather than “retrain everything”).

  2. Non-stationarity requires explicit drift and decay handling, not ad-hoc updates. Practice: continual learning emphasizes non-stationarity as a first-class maintenance condition. Source: Parisi et al., Continual Lifelong Learning with Neural Networks (arXiv:1802.07569, 2019); De Lange et al., A Continual Learning Survey (arXiv:1909.08383, 2021). Alignment: B.3.4 supplies use-qualified currentness. G.11 interprets changed conditions and justified review signals before planning an affected-scope refresh; elapsed time alone implies neither truth decay nor deprecation. Adoption: Adapt (refresh of conceptual artefacts and evidence closures, not untracked model mutation).

  3. Quality-Diversity requires archive semantics and comparability under descriptor and distance evolution. Practice: QD methods treat the archive as the primary result and track changes under policy and edition conditions. Source: contemporary QD families such as CMA-MAE (arXiv:2205.10752) and differentiable QD (arXiv:2106.03894). Alignment: QD-specific meaning lives with the subject patterns; G.11:Ext.QDRefreshWiring ensures edition pins and scope pins exist so targeted archive refresh is admissible. Adoption: Adopt (set and archive preservation; no covert scalarization).

  4. Open-endedness co-evolves environments and agents; transfer rules must be versioned. Practice: POET-class open-ended systems require explicit transfer rules and environment validity constraints. Source: Wang et al., POET (arXiv:1901.01753, 2019); later generator-family claims require a named G.2 SoTA pack or exact current source. Alignment: G.11:Ext.OEERefreshWiring requires TransferRulesRef.edition and scope pins so refresh reruns remain comparable and auditable. Adoption: Adopt and adapt (adapted to Part G pin and UTS publication discipline).

  5. Efficient orchestration benefits from bandit and early-stopping scheduling, but scheduling must not redefine trigger, action, parity, shipping, or Part-G-wide default semantics. Practice: modern hyperparameter and experiment scheduling uses bandit-style resource allocation and asynchronous early stopping. Source: ASHA (arXiv:1810.05934) and BOHB (arXiv:1807.01774) as representative post-2015 scheduling practice. Alignment: scheduling is expressed as RefreshQueue and RefreshPlan@Context policy pins (RefreshPriorityPolicyIdRef, BudgetDeclRef) so core semantics remain stable and the exact U.WorkPlan stays separate from dated Work. Adoption: Adapt (useful practice, but quarantined outside core norms).