Jump to text
On this page
Copy and source
Published source confirmed at last check

Source changed 2026-11-18 21:11:39 UTC · snapshot created 2026-11-18 21:12:35 UTC · last check 2026-11-18 22:00:10 UTC

C.22.1 - Task-family adaptation signature

Status: Stable

Purpose. Compare how a holder, dyad or bounded specialist portfolio acquires usable specialization on a declared task family. An adaptation signature connects the threshold, acquisition time and cost, and prior exposure with any claimed transfer, retention, downside or entry into a new solution corridor. An explicit account in ordinary prose is sufficient when these distinctions and their grounds are recoverable; a shared record can support later selector or parity use.

Keywords. adaptation signature; task-family specialization; time-to-threshold; budget-to-threshold; prior exposure; corridor entry; stepping stone; transfer; retention; downside field.

C.22.1:1 - Problem frame

Final task score alone does not tell whether a holder, dyad, or bounded specialist portfolio acquired usable specialization quickly, under what budget, with what prior exposure, whether the resulting competence transferred, or whether the adaptation path entered a new solution corridor. If those elements are not published together, the adaptation claim splinters across task typing, probe notes, selector prose, and parity notes, and later readers can no longer tell what was being compared.

C.22.1:2 - Problem

A comparison of specialization acquisition needs a common task family and work target, with enough acquisition history to distinguish equal final scores reached under different conditions.

C.22.1:3 - Use this when

  • the live task-family adaptation claim is not only that a holder or dyad solved a task, but how fast it acquired usable specialization on a declared task family
  • comparison must account for the work-measure threshold target, prior exposure, adaptation budget, transfer field, and reuse window
  • movement into a new solution corridor or stepping-stone family is part of the novelty claim

C.22.1:4 - What goes wrong if missed

  • adaptation claims collapse into vague got better language with no declared work-measure threshold target or budget-to-threshold account
  • parity later compares outcomes that were reached under different prior exposure, different work-measure threshold targets, or different reuse windows
  • nonhuman or unfamiliar solution corridors are either treated as novelty or dismissed as noise because the corridor entry was never typed

C.22.1:5 - What this buys

  • adaptation speed becomes reviewable by value on the same declared TaskFamily and work target
  • later G.5 / G.9 portfolio and parity work can compare specialization claims on the same declared basis
  • stepping-stone or solution-corridor movement becomes visible as one typed part of the adaptation claim

C.22.1:6 - Forces

ForceTension
Threshold crossing vs final scoreA static outcome can look similar even when one system specialized much faster or more cheaply than another.
Local novelty vs reproducible evidenceCorridor-entry claims matter, but they are easy to overstate when no baseline or entry evidence is published.
Task anchor vs adaptation-signature questionTask typing fixes what is being acquired; selector and parity rules govern subsequent comparisons.
Reuse upside vs specialization costTransfer, retention, and downside matter to the same claim even when the first threshold crossing looks impressive.

C.22.1:7 - Solution — one adaptation signature over the C.22 anchor

  • Use one shared adaptation-signature field set for this question. G.5, G.9, and later notes may cite or consume it, but they should not silently rename threshold, prior-exposure, transfer, downside, or corridor-entry terms.
  • Bind the adaptation signature to the declared TaskFamilyRef or TaskSignature and work target. Name whose capability is being judged and what tools, references or help are part of the intended work. Record who performed the operation on which the threshold depends and what help was supplied during practice and the check. A joint performance can establish what the participants can do together. For a claim about one participant, use a performance that shows what that participant can do under the intended support conditions. Reuse an existing account when it supplies those grounds.
  • The signature should expose at least:
    • thresholdTarget
    • timeToThreshold
    • budgetToThreshold
    • postThresholdEfficiency?
    • priorExposureDeclaration
    • transferTarget?
    • transferGain?
    • retentionWindow?
    • downsideEffect?
    • corridorEntryBaseline?
    • corridorEntryEvidence?
    • steppingStoneEvidence?
  • Anchor these fields to the same work target and work-measure threshold semantics declared by C.22. State the time origin, budget boundary and threshold-observation rule. If the threshold was not reached or the available history cannot establish its first crossing, say so and retain the observed limit; do not invent a crossing time or cost.
  • C.22 carries the declared task-family anchor, task typing, and baseline TaskSignature. C.22.1 narrows the adaptation-signature question to threshold timing, reuse, downside, and corridor-entry disclosure over that existing anchor.

C.22.1:8 - Corridor, transfer, and durability discipline

  • If the adaptation claim depends on entering a new solution corridor, publish the corridorEntryBaseline first: the prior repertoire, baseline set, or comparison family relative to which corridor entry is being claimed.
  • State corridorEntryEvidence that supports the claimed difference from that baseline, such as a reproducible solution class, a stable descriptor shift or a supported stepping-stone sequence. Name what distinguishes the corridor; an unfamiliar sample alone does not establish entry.
  • If a stepping stone mattered, publish the stepping-stone evidence as part of the adaptation signature.
  • Corridor or stepping-stone notes do not replace the work-measure threshold account; they explain why the adaptation path matters, not whether the threshold was reached.
  • A fast threshold result is not yet enough to claim durable specialization.
  • For transfer, identify the target task family, baseline and observed change there. To attribute that change to the earlier practice, give grounds that distinguish its effect from repeat testing, practice on the target or changed assistance when those explanations are plausible (C.28). State how far the evidence reaches across the target family.
  • For retention, report when and under what support conditions performance was demonstrated again. Use a comparable criterion over the stated delay. An unobserved interval can remain open.
  • If specialization harms neighboring task families, narrows reusable competence, or creates de-specialization cost, publish that in downsideEffect?.
  • If post-threshold performance matters to later exploitation, publish postThresholdEfficiency? to show performance after threshold crossing.

C.22.1:9 - Worked moment

Consider two research setups, A and B, adapting to the same catalyst-search family. This is an illustrative comparison. The declared threshold is at least 80 accepted predictions out of the same 100 held-out cases under a fixed validation rule. Both start with the same training corpus and no prior exposure to those held-out cases. Time starts when adaptation begins; the budget counts all adaptation and evaluation calls. The rule checks threshold attainment at the end of each round.

A first passes a scheduled check after 2 hours and 120 calls, B after 6 hours and 480 calls. Both finish at 85/100. The histories show what each setup spent by its first observed pass. To conclude that either acquired the capability during adaptation, establish how it performed beforehand under comparable conditions. A failed check followed by a pass brackets the observed change at the declared checkpoints.

On a named adjacent family, a before/after test gives A 60/100 then 70/100 and B 60/100 then 60/100. A improved by ten percentage points on this test. B’s observed score stayed the same. These observations leave open how much change the original practice caused. A comparison that accounts for repeat testing, other training and assistance can support a transfer claim for the conditions it covers. Retention beyond the measured run is untested. A corridor-entry claim would need the prior repertoire and evidence of a repeatable solution class absent from it.

The comparison can finish with these bounded claims. If A’s prior exposure later proves different, reopen the equal-exposure comparison while retaining the observed histories. A later retention claim needs its own window and observations. A selector may use the signature under its applicable rule; the signature alone chooses neither setup.

C.22.1:10 - Consequences

  • Acquisition time and cost can be compared with prior exposure and any claimed post-threshold efficiency on one recoverable basis.
  • Selector and parity pattern applications can consume a stable upstream adaptation signature without silently redefining its fields.
  • Claimed corridor entry and known downside remain visible alongside the specialization gain.

C.22.1:11 - Rationale

The reader needs one place where the adaptation claim stays whole. C.22 keeps the task family and work target explicit.

A.15 distinguishes probe plans from performed Work. A claim about a particular dated Work occurrence needs A.13 and independent A.15.1 admission; precise assignment-bound attribution additionally uses F.6. C.24 defines call-planning CheckpointReturn records after an action or option is fixed; E.16 governs budget, guard, and ledger enforcement for autonomy claims. Probe results, checkpoint records, and budget records may supply evidence for the adaptation claim.

G.5 and G.9 later compare several candidates or parity runs. C.22.1 keeps the specialization story readable across those neighbouring pattern applications by connecting threshold timing to the claimed reuse, downside and corridor-entry evidence.

C.22.1:12 - SoTA-Echoing

Claim 1. A comparison of specialization acquisition keeps the declared work-measure threshold and prior exposure explicit. It reports the observed time and cost to threshold. If the first crossing is unresolved, give the time or cost bounds the observations support. Distinguish a successful first check at the end of a month from evidence that acquisition took a month. Report non-attainment for the monitored interval only when the observation rule supports it.

Practice source, local alignment, and adoption decision. FactorMiner v1 (2026-02-16), §§4.1–4.4, separates factor quality and redundancy, the effect of experience memory on screening yield, and factor-evaluation computation time. SkillOpt v2 (2026-05-25), §§3.1, 3.4–3.5 and 4.3–4.4, changes an external skill text while holding the task-execution model fixed, selects text revisions through held-out validation, and evaluates transfer and training tokens per test-score gain. This pattern adopts these distinctions among the changed object, attained result, evaluation or training cost, and tested reuse.

The shared field set is an original FPF construction. For the question in §§1–6, §7 adapts those narrower comparisons by connecting the acquisition history to one TaskFamilyRef or TaskSignature and work-measure threshold. A terminal score and total cost per score gain can describe the attained result and average improvement cost, yet leave the resources needed to reach a particular threshold unresolved.

Recording threshold timing and prior exposure adds observation effort but makes histories with the same final score distinguishable; §8 then separates claimed transfer, retention, and downside. The pattern rejects substituting either paper’s reported metrics for this shared acquisition account.

Reopen the comparison if a rival protocol supports the same specialization judgement with less observation, or exposes a consequential difference that these fields cannot express.

Claim 2. A corridor-entry claim needs a baseline and evidence for the claimed difference.

C.17 separates novelty and diversity relative to a declared comparison basis; C.18 distinguishes generated candidates, retained archives and fronts. Here that distinction supplies a conditional question: when acquisition includes a claimed new corridor, which prior repertoire and observations establish it? The fields in §8 make that evidence recoverable without treating one unfamiliar sample as reproducible entry.

Claim 3. Selector and parity comparisons depend on compatible interpretations of the fields they consume.

G.5 and G.9 consume this shared field set. Their users must retain its threshold, prior-exposure and conditional reuse meanings when comparing candidates; a silent downstream redefinition would change the comparison basis.

Evidence-source note. Use each source result with its measured object and evaluation conditions. Factor admission and skill-edition selection concern different objects from a holder’s acquisition of usable specialization. For timeToThreshold and budgetToThreshold, relate the acquisition evidence named in §11 to the declared threshold and prior exposure; faster evaluation or tokens per score-point gain alone cannot supply that history. Transfer, retention, and corridor-entry claims keep their separately stated evidence requirements.

Source-bound anchor familySource-use relation/currentnessWhat it disciplines in this pattern
Novelty and search accountsC.17 and C.18 provide the internal distinctions used above.Corridor baseline, descriptor shift, stepping-stone evidence, and whether novelty is reproducible rather than one unfamiliar sample.
Agentic adaptation studiesFactorMiner v1 and SkillOpt v2: versioned preprints, used for the bounded comparisons above.FactorMiner distinguishes factor quality, memory-guided yield, and evaluation cost; SkillOpt distinguishes text revision, validation, transfer, and training cost. The common acquisition field set is constructed in §§7–11.
Transfer / retention evaluationThe applying comparison must name its observations, task family and evaluation conditions.Transfer target, retention window, downside and reuse evidence remain separate from first threshold attainment.

C.22.1:13 - Relations

C.27 temporal-claim relation.

  • C.27 may flag: a claim that a holder, dyad, team, explicitly scoped specialist portfolio, or agent acquires usable specialization faster on one declared TaskFamilyRef or TaskSignature. A method may be selected, refined, or retired as part of that story.

  • This pattern keeps: threshold target, time-to-threshold, budget-to-threshold, prior exposure, transfer, retention, downside, corridor-entry evidence, and adaptation-signature fields.

  • “learns faster” wording without a task-family anchor leaves the adaptation claim incomplete. Faster threshold crossing alone does not establish durable specialization; retention needs its own window and evidence, and any additional transfer, downside or corridor-entry claim needs its stated grounds.

  • Next-question boundary: classify the claim as Dyn1 when only a trend is live; use C.24 for call planning after the action or option is fixed; use the C.22.1 adaptation signature when specialization is the live adaptation question.

Builds on: C.22 TaskSignature anchoring, C.19.1 BLP compatibility, A.15 system-role-kind and assignment, Method, WorkPlan, and Work-occurrence separation, C.24 call planning and CheckpointReturn semantics after the action or option is fixed, E.16 budget enforcement.

Coordinates with: G.5 selector specialization profiles, G.9 adaptation parity, G.11 later telemetry/refresh reuse.

Coordinates with: E.23 when repeated object improvement includes a claim of durable task-family specialization. C.22.1 supplies the adaptation account for that claim; E.23 governs the improvement loop.

Constrained by: E.10 lexical discipline.

C.22.1:14 - Not this pattern when

  • the claim only needs to name the task family and work-measure threshold target, with no adaptation-speed or transfer claim; ordinary C.22 anchoring is enough
  • the question under repair is already selector or parity law across candidate selected sets; that belongs to G.5 / G.9
  • the threshold or prior-exposure basis cannot yet be established: retain a narrower observation or an unresolved adaptation claim. Missing corridor evidence leaves that corridor claim unresolved; it need not prevent a separately supported threshold account

C.22.1:15 - Conformance checklist

  • CC-C22.1-1 An adaptation signature SHALL bind to one declared TaskFamily or TaskSignature, one work target, and one work-measure threshold target.
  • CC-C22.1-2 An adaptation signature SHALL report prior exposure and the observed time and budget to threshold, or the supported limits on each. Distinguish a known crossing, a threshold met by the check with the first crossing undetermined, and a threshold known to have remained unmet. If the observations establish none of these, report what remains unknown.
  • CC-C22.1-3 Any declared transfer, retention, post-threshold-efficiency, downside, corridor-entry, or stepping-stone claim SHALL identify what is claimed and for which target. Give the grounds for that claim under §8, including the baseline or observation window when the claim depends on it.
  • CC-C22.1-4 This pattern may refine specialization timing and reuse claims over the declared C.22 anchor, but it SHALL NOT redefine acceptance-gate thresholds, task-family attachment, or selector/parity law governed by another FPF pattern.
  • CC-C22.1-5 Downstream selector/parity pattern applications SHALL cite or consume the same published adaptation-signature field set rather than silently redefining threshold, prior-exposure, transfer, retention, downside, or corridor-entry terms.

C.22.1:End

Referenced in the corpus

21 literal mentions in other sections. Read their context to establish the relation.