Claim Status, Freshness, and Research Handoffs

A scientific claim does not become truer or falser merely because time passes. What can change is the evidence available for a precisely fixed claim, the strength of the conclusion that evidence licenses, and whether the assessment has been checked after a correction or new result. A useful learning page therefore preserves reproducible mathematics, dates its scientific assessment, and says exactly which later events require a new assessment rather than silently replacing the old one.

Required background. Evidence Independence, Circularity, and Double Counting determines which sources provide genuinely distinct support. Falsifiers, Negative Results, and Counterexamples determines whether a negative result challenges the claim, its assumptions, or only one method.

Helpful background. Claim–Evidence Records, Replication, and Retraction Handling supplies the general update model.

Evidence cutoff. The source review on this page includes versions available through 27 August 2026. The next scheduled review is no later than 27 February 2027, or sooner if a trigger named below occurs.

Reading path. First separate claim identity from status, then compare the three assessment packets. The worked successor update shows what a trigger changes; the handoff and exercises turn the method into practice.

A claim has an identity before it has a status

Freeze the proposition before asking whether new evidence changes its standing. Write a claim version $P_v$ with enough information to determine what would count as agreement or failure:

• theory or dictionary, state or ensemble, observable, and physical domain;
• boundary conditions, conventions, regulator, and order of limits;
• approximation order and the norm or uncertainty convention;
• source, method, and data versions that define the comparison.

Changing one of these ingredients may create a successor proposition $P_{v+1}$ rather than new evidence about $P_v$. For example, exact recovery on an entire finite-$N$ Hilbert space and approximate recovery on a low-energy code subspace are not two confidence levels for one sentence. They are different claims.

Let an assessment snapshot be

$$A_k=\mathcal A\!\left(P_v\mid E_{\leq c_k},M_k\right),$$

where $c_k$ is the evidence cutoff, $E_{\leq c_k}$ is the frozen evidence packet, and $M_k$ contains the analysis method and assumptions. Preserve $A_k$ when evidence changes and append $A_{k+1}$. This makes it possible to reproduce what was known at either cutoff and to explain every upgrade or downgrade.

Five notions must remain separate.

Independent dimensions of a scientific claim record

Dimension	Question it answers	What a change means
Claim identity	What exact proposition, observable, domain, and approximation are being assessed?	A changed meaning, theory, observable, or domain forks a new proposition.
Truth	Is that fixed proposition actually true?	Time does not alter the truth value; it can alter what we are justified in believing.
Epistemic status	Is the claim proved under hypotheses, controlled asymptotically, numerically supported, empirically bounded, conjectural, contested, or excluded in a stated domain?	New relevant evidence can raise or lower the licensed claim ceiling.
Freshness	Has the assessment been reviewed through its cutoff and after every declared trigger?	Missed review makes the assessment stale, not automatically false.
Source and reproducibility	Are sources current, corrected, superseded, or withdrawn, and can the stated calculation be reproduced?	A source event prompts impact analysis; it does not by itself prove or negate the claim.

No single badge should collapse these dimensions. “Current” does not mean “correct,” and “stable” does not mean “immutable.”

What the record must preserve

A compact record should let another researcher recover both the proposition and the reason for its present status. It needs:

1. the exact claim sentence, observable, estimator, domain, conventions, and control parameters;
2. statistical, systematic, truncation, regulator, and model uncertainties in the convention actually used;
3. exact source versions—version of record or arXiv revision, dataset release, code commit or archive, environment, and checksums when byte identity matters;
4. supporting and contrary evidence grouped by dependence, with an edge from every conclusion to the sources and methods it uses;
5. scientific status, publication status, and reproducibility status as separate fields;
6. evidence cutoff, assessment date, next review date, and event-driven triggers;
7. predecessor and successor links, the reason for each change, and the strongest subclaim that survives; and
8. a concrete continuation: the unresolved proposition, missing assumption or test, required calculation or data, and a success or failure criterion.

A reusable status sentence is:

Assessed through [cutoff], claim $P_v$ in domain $D$ is supported only to [ceiling] by [independent evidence]; [contrary result] limits it, [weaker claim] survives, and [event] triggers reassessment.

The clock has both a scheduled and an event-driven part. If $T(P_v)$ is the declared trigger set and $U_{(t_k,t]}$ the events since the last review, then

$$\text{review due} =\left(t\geq t_{\mathrm{next}}\right) \lor\left(T(P_v)\cap U_{(t_k,t]}\neq\varnothing\right).$$

The interval is a risk-management choice, not a scientific half-life. Corrections, retractions, superseding data, changed likelihoods or calibration, failed reproductions, and new same-domain counterexamples should normally trigger an earlier review.

The vocabulary for reproducibility also matters. Analytic verification checks a derivation from frozen assumptions. Computational reproducibility reruns the same data, code, method, and conditions. Independent reimplementation uses new code on the same data. Replication uses independently obtained data to address the same scientific question. Robustness varies regulators, priors, models, or methods. The National Academies distinguishes reproducibility from replication in this way; neither one alone proves physical correctness (NASEM 2019, Chapter 3, pp. 43–47).

Three assessment packets with three clocks

The following records show why one review cadence cannot serve every statement. Compound subjects are split into atomic propositions inside each packet. The gravitational-wave example uses the fourth Gravitational-Wave Transient Catalog (GWTC-4.0) and its LIGO–Virgo–KAGRA (LVK) analysis; MDR means modified dispersion relation. In the reconstruction example, CCKLP abbreviates Cao–Cheng–Karthikeyan–Li–Preskill.

Three assessment packets at the 27 August 2026 evidence cutoff

Field	Euclidean GKPW	Finite-N reconstruction	GWTC-4.0 modified dispersion
Atomic claims and observables	GKPW-D: for a specified dual pair, the CFT generating functional equals the bulk path integral. GKPW-S: in a semiclassical regime, derivatives of the saddle approximation to log Z generate connected boundary correlators.	FN-A: approximate wedge recovery is controlled in the CCKLP perturbed-code model. FN-C: an ordinary finite-N CFT has one nontrivial region-independent logical algebra with code-preserving regional representatives.	GW-MDR: no statistically significant evidence for a non-GR dispersive phase was found in the tested templates. GW-mg: the massive-graviton specialization yields an upper bound on mg.
Domain and conventions	Euclidean AdS/CFT pair, fixed source sign and boundary data, renormalized action, chosen saddle, weak bulk coupling, and controlled higher-derivative expansion.	Named code projector, boundary region and complement, dressed operator, center convention, recovery channel, error norm, and order of the N, regulator, and code-size limits.	GWTC-4.0 O1–O4a compact-binary signals selected by the collaboration; the dispersion, waveform, cosmology, calibration, likelihood, and prior conventions of §3.1 of the cited analysis.
Controls and uncertainty	Bulk loops, string corrections, competing saddles, counterterm scheme, and contact terms. Separated-point nonlocal structure is more stable than local terms.	Leakage or relative-entropy error, code dimension, gravitational dressing, complementary-region detectability, center or superselection sectors, and model-to-CFT transfer.	Detector calibration, noise, waveform family, source parameters, cosmology, event selection, prior dependence, and the stated 90% Bayesian credibility convention.
Frozen sources	Gubser–Klebanov–Polyakov arXiv:hep-th/9802109v2 and Witten arXiv:hep-th/9802150v2, checked against their journal records.	Cao et al. arXiv:2603.13475v2; Witten arXiv:2606.18639v3; Terashima arXiv:2607.08684v1.	LVK arXiv:2603.19020v2; LIGO-P2600129-v2; MDR.tar.gz; posterior release 10.5281/zenodo.21403342; and environment-igwn-py310-20241106.yml for the International Gravitational-Wave Observatory Network software environment.
Support and dependence	Gubser–Klebanov–Polyakov and Witten are contemporaneous formulations of the same dictionary program, not two independent confirmations. Analytic differentiation verifies GKPW-S once its premises are frozen.	Cao et al. define the model and Witten analyzes that framework, so their support forms one dependent chain. Terashima supplies a partly independent argument about the stronger FN-C proposition.	The paper, posterior samples, scripts, calibration inputs, and catalog are components of one LVK analysis packet. They enable checks of one inference but are not independent empirical confirmations.
Scientific status and ceiling	GKPW-D is a proposed dictionary entry for specified dual pairs. GKPW-S is a controlled computational approximation in semiclassical regimes, not a theorem of generic holography.	FN-A is model-supported under explicit assumptions. FN-C is actively contested and is neither established nor universally refuted.	GW-MDR is a null result in named templates; GW-mg gives mg ≤ 1.92 × 10−23 eV/c2 at 90% credibility. Neither identifies a microscopic quantum-gravity theory.
Publication and reproducibility	Both foundational sources have journal versions. GKPW-S is analytically verifiable from frozen conventions; that is not an experimental replication of GKPW-D.	All three 2026 sources are versioned preprints at this cutoff. Their analytic arguments can be checked, but no same-domain finite-N CFT calculation independently adjudicates FN-C.	The collaboration preprint and data release are versioned. Public posteriors and scripts support reproduction of released plots and combinations; this page has not independently rerun event-level parameter estimation.
Dated contrary-source check	No correction, withdrawal, or same-domain refutation was identified in the frozen arXiv and journal records through the cutoff. Loop, string, saddle, and contour effects are declared limitations rather than contrary sources.	Terashima challenges FN-C under locality and detectability assumptions. It does not target FN-A under identical premises, so it is not counterevidence to that model result.	No correction affecting the MDR archive was identified in the paper and LIGO-P2600129-v2 records through the cutoff. Waveform and calibration alternatives are failure modes, not observed contrary results.
Triggers and strongest survivor	A source correction, competing saddle, or changed contour triggers review. The source-to-correlator prescription and GKPW-S survive only where their frozen premises remain valid.	A proof, counterexample, matched CFT construction, or changed error estimate triggers review. The CCKLP model result survives; Terashima's argument leaves region-adapted reconstruction as a possible, separately unproved interpretation.	A corrected MDR archive, changed calibration or waveform analysis, or stronger same-template bound triggers review. Until then, the released template-dependent bound survives.
Review clock and continuation	Review on a source correction, a changed canonical derivation, or a change to the page's conventions; otherwise the six-month page review is sufficient.	Review immediately after a proof, counterexample, revised preprint, matched CFT construction, or changed error estimate. Compare the same algebra, code, dressing, and norm.	Review after a corrected MDR archive, changed calibration or waveform analysis, new catalog rerun, or materially stronger same-template bound. Recompute before claiming supersession.

GKPW: separate the dictionary from the saddle

In Euclidean signature the two steps are

$$Z_{\mathrm{CFT}}[J] =Z_{\mathrm{bulk}}[\phi_{(0)}=J], \qquad Z_{\mathrm{bulk}}[J] \simeq \exp\!\left[-I^{\mathrm{ren}}_{E,\mathrm{on\text{-}shell}}[J]\right].$$

The first equality states the proposed dictionary for a specified dual pair. The second evaluates the bulk path integral by a chosen classical saddle. Boundary conditions, counterterms, saddle dominance, and approximation order belong to the second claim; a Lorentzian formula would additionally require a state and contour prescription. Derivatives of $Z_{\mathrm{CFT}}[J]$ generate full correlators, whereas derivatives of $W[J]\equiv\log Z_{\mathrm{CFT}}[J]\simeq-I^{\mathrm{ren}}_{E,\mathrm{on\text{-}shell}}[J]$ generate connected correlators in the fixed source-sign convention. The full GKPW derivation carries out this step. The claim ceiling follows directly from the distinction emphasized by Witten 1998, §2.3 and from holographic renormalization (de Haro, Skenderis, and Solodukhin 2001, §5.1).

This record is comparatively durable because its assumptions can be frozen and the calculation analytically verified. Durability is not exactness: a source correction that changes a sign, counterterm, or saddle premise still requires impact analysis.

Finite N: compare like with like

The 2026 papers do not supply a vote on one proposition. Cao et al. study approximate subsystem erasure-correcting codes obtained by perturbing exact codes and define a state-dependent proto-area through optimal recovery. In that framework, Witten shows that under the stated dimension and scaling assumptions the recovery correction can be exponentially small in $1/G$ relative to an area-function correction. This is a quantitative model result, not a universal finite-$N$ CFT theorem (Cao et al. 2026v2, abstract and §§2–4; Witten 2026v3, §§3.7–3.9).

Terashima studies a stronger target: one region-independent logical operator with code-preserving representatives in several regions of an ordinary finite-$N$ CFT. The argument uses locality plus an additional detectability premise and leaves room for center, topological, or superselection data. It does not rule out region-adapted entanglement-wedge reconstruction; it leaves that as a possible, separately unproved interpretation while challenging the stronger common-algebra claim (Terashima 2026v1, §§2.1–3.2 and §§4–5).

Accordingly, the dated status is “active, versioned arguments about partly non-equivalent propositions.” Kibe, Mandayam, and Mukhopadhyay 2022, §§3.4 and 4 provides the specialist review baseline for the pre-2026 QEC interpretation; the three versioned primary sources are needed for the later dispute. The detailed evidence comparison and the finite-$N$ reconstruction limits own the physics. This page records why a status update must first normalize the algebra, code, dressing, norm, and limit.

GWTC-4.0: a bound is an analysis packet

The phenomenological ansatz includes

$$E^2=p^2c^2+A_\alpha p^\alpha c^\alpha,$$

which produces a frequency-dependent propagation phase in the detector waveform. The massive-graviton specialization is $\alpha=0$ with $A_0=m_g^2c^4$. The LVK analysis combines selected O1–O4a events with event-level likelihoods and explicit waveform, calibration, source, prior, and cosmological assumptions. It found no statistically significant evidence for a non-GR dispersive phase and reported

$$m_g\leq1.92\times10^{-23}\,\mathrm{eV}/c^2 \qquad\text{at 90\% credibility}.$$

The paper version alone is not the whole result. The frozen packet includes LVK 2026v2, §3.1, eqs. (11)–(13), LIGO-P2600129-v2, MDR.tar.gz, the posterior release, and environment-igwn-py310-20241106.yml. The release note says that v2 corrected products from the Test Infrastructure for GEneral Relativity (TIGER), which belongs to another analysis; MDR.tar.gz is a separate dependency. The correct response is to inspect the dependency edges, not to mark the graviton-mass bound changed merely because the release acquired a new version. The gravitational-wave propagation page develops the physical inference.

A later source changes only dependent conclusions

Consider an intentionally earlier local snapshot with cutoff 30 June 2026. It contains Cao et al. v2 and Witten v1, but not Terashima v1 of 9 July or Witten v3 of 10 July. Terashima v1 is a declared same-topic trigger: on 9 July the finite-$N$ assessment enters review required since 9 July 2026, and its Research continuation becomes stale until the new argument is compared. Witten v3 triggers a source-version check the next day. The stale label reports an uncompleted assessment; it does not say that either proposition is false.

The completed update proceeds as follows.

1. Identify the events. Terashima introduces an interpretive challenge; Witten v3 supersedes the frozen source version but labels v2 and v3 as minor corrections.
2. Normalize the propositions. Separate approximate recovery in the CCKLP perturbed-code model from a shared region-independent logical algebra in an ordinary finite-$N$ CFT.
3. Verify the frozen calculation. Witten v1 already contains the quantitative recovery-versus-area hierarchy, and the version comparison finds no stated change to that conclusion in v3. The frozen GKPW source differentiation also re-evaluates unchanged because neither July source enters its premises.
4. Trace dependencies. The finite-$N$ assessment depends on the new papers. The GKPW derivation and GWTC-4.0 bound do not.
5. Run impact analysis. Retain only derivations and subclaims whose premises and sources are unaffected; do not assume that every equation survives merely because it still renders.
6. Append a successor. Preserve the 30 June snapshot, record “reviewed, no material version impact” for Witten v3, and record Terashima’s target and assumptions as the reason FN-C becomes contested.
7. State the survivor and next test. Preserve FN-A in its model. Terashima’s argument leaves region-adapted reconstruction possible but separately unproved. The Research continuation must compare a matched algebra, code, dressing, detectability assumption, and error norm.

A post-cutoff update without silent overwriting

Item	30 June 2026 snapshot	Post-cutoff event	27 August 2026 successor
Freshness state	Reviewed through 30 June; the finite-N packet has no post-cutoff source.	Terashima v1 triggers “review required since 9 July”; Witten v3 adds a version check on 10 July. The associated Research continuation is stale pending comparison.	Reviewed successor dated 27 August; the prior snapshot remains reproducible and linked rather than overwritten.
Approximate-code proposition	Supported within the CCKLP perturbed-code construction; Witten v1 already gives the quantitative hierarchy under its stated assumptions.	Witten v2 and v3 make minor corrections; compare versions to determine whether a premise, formula, or conclusion changes.	Still model-supported; the comparison finds no stated material impact, so the successor uses v3 and records “reviewed, no material version impact.”
Shared finite-N CFT algebra	Not established by the model calculation; no same-packet negative argument included.	Terashima v1 challenges it using locality and detectability for ordinary supergravity sectors.	Downgraded to actively contested, not refuted universally: the argument targets the stronger shared algebra and has stated exclusions.
Unaffected claims	GKPW saddle calculation and GWTC-4.0 bound.	No dependency edge reaches either record.	Both retain their prior scientific status; only their ordinary review clocks continue.
Research continuation	Compare recovery estimates with a concrete holographic realization.	A sharper algebraic target becomes available.	Construct or exclude the same dressed operator in the same finite-N code using both recovery and commutant tests, with one declared norm; test region-adapted reconstruction separately.

This is also the correct pattern for a correction, supersession, withdrawal, or retraction. A correction may have no impact, local impact, or claim-changing impact. Superseded means a newer packet is preferred, not necessarily that the old result was wrong. Withdrawal or retraction removes a source as licensed support; it does not prove the opposite proposition. Refutation requires contrary evidence or a counterexample addressing the same fixed proposition and domain. Publication metadata such as Crossmark can announce a source event, but it cannot perform the scientific impact analysis (COPE 2019, Retraction Guidelines, pp. 3–5; Crossref, “Crossmark”).

From a learning page to a research question

The division is scientific, not merely administrative.

• Learn keeps definitions, checked derivations, benchmark calculations, and a dated statement of their strongest licensed interpretation. If a premise is affected, Learn must change too.
• Research tracks evolving comparisons, open alternatives, unresolved assumptions, and proposed discriminating tests. The Holography and Quantum Gravity research guide has its own evidence cutoff, which readers should check rather than assuming it includes every source reviewed here.
• Corrections should report the exact page, statement, source, and suspected impact through the functional Feedback and corrections route so the editorial review can begin.

A useful handoff does not say only “more work is needed.” For the finite-$N$ example it names the open proposition, positive and negative results, disputed detectability premise, candidate dressed algebra, error norm, cutoff, and a decisive criterion. The existing question on bulk reconstruction beyond semiclassical code subspaces is the natural continuation. The next Learn chapter, Large-N Holographic Evidence, studies how the asymptotic control behind many such statements is actually established.

Common pitfalls

Treating a newer source as contrary evidence. It is contrary only if it addresses the same proposition and domain. A paper about a stronger algebra can coexist with a positive theorem for a narrower approximate code.

Confusing source status with scientific status. Peer review does not prove a claim, and preprint status does not make it false. A retraction removes support but does not establish the negation.

Calling every rerun a replication. Same-data, same-code reruns test computational reproducibility. New code on the same data is an independent implementation; independently acquired data are needed for replication of the scientific result.

Letting persistent identifiers stand in for validation. FAIR metadata can make evidence findable and reusable, while a versioned identifier and checksum can bind a record to particular bytes. None establishes that the inference is correct (Wilkinson et al. 2016, Box 2, Principles F1–R1).

Automatically superseding a bound. A newer catalog does not replace a derived bound until the relevant analysis is rerun or a justified transfer calculation is supplied.

Exercises

Two clocks. A GKPW derivation was checked today, but its cited review article is five years old. Is the calculation stale?

Solution: separate the clocks

Not from the article’s age alone. Record the exact primary sources and verify the derivation under the page’s conventions. The interpretive literature review has a separate clock and may be stale even when the calculation remains analytically verified.

A corrected release. A public data release is revised because a plotting script in an unrelated analysis had a bug. What should happen to a bound that depends on a separate archive?

Solution: trace the corrected release

Mark the source event and inspect dependency edges. If the bound uses neither the corrected script nor its outputs, record “reviewed, no impact” and preserve the bound. If an edge is affected, rerun the relevant analysis before issuing a successor conclusion.

A retracted benchmark. A numerical benchmark used to calibrate one finite-$N$ recovery estimate is retracted, but the analytic code theorem does not use it. What survives?

Solution: preserve the unaffected theorem

Withdraw the benchmark as support and reassess every numerical conclusion that depends on it. Preserve the analytic theorem with its hypotheses, along with any independently supported subclaims. Retraction of the benchmark does not refute the theorem.

Capstone handoff. Write a one-sentence successor status for the finite-$N$ dispute after the July 2026 papers.

Solution: write the successor status

Assessed through 27 August 2026, approximate recovery is quantitatively supported in the specified CCKLP perturbed-code framework, while a shared region-independent logical algebra for ordinary finite-$N$ CFT supergravity sectors remains contested by a locality-and-detectability argument; region-adapted reconstruction remains possible but separately unproved, and a matched algebra, dressing, code, and error-norm comparison is the next discriminating test.

Chapter comparison tools

Use the chapter structure diagram to trace a proposition from dictionary and regime through observable and evidence. The validity and failure diagram identifies which premise a later event can affect. The claim-domain table keeps the licensed claim ceiling beside its failure conditions.

References

• Cao, ChunJun, Gong Cheng, Krishnanand Karthikeyan, Cathy Li, and John Preskill. 2026. “State-Dependent Geometries from Magic-Enriched Quantum Codes.” arXiv:2603.13475v2, 27 June 2026. Versioned preprint.
• Committee on Publication Ethics. 2019. Retraction Guidelines, version 2. PDF.
• Crossref. “Crossmark.” Service description.
• de Haro, Sebastian, Kostas Skenderis, and Sergey N. Solodukhin. 2001. “Holographic Reconstruction of Spacetime and Renormalization in the AdS/CFT Correspondence.” Communications in Mathematical Physics 217, 595–622. DOI; Open preprint.
• Gubser, Steven S., Igor R. Klebanov, and Alexander M. Polyakov. 1998. “Gauge Theory Correlators from Non-Critical String Theory.” Physics Letters B 428, 105–114. DOI; arXiv v2.
• Kibe, Tanay, Prabha Mandayam, and Ayan Mukhopadhyay. 2022. “Holographic Spacetime, Black Holes and Quantum Error Correcting Codes: A Review.” European Physical Journal C 82, 463. DOI; Open preprint.
• LIGO Scientific Collaboration, Virgo Collaboration, and KAGRA Collaboration. 2026. “GWTC-4.0: Tests of General Relativity. II. Parameterized Tests.” arXiv:2603.19020v2, 20 July 2026. Versioned preprint; versioned data and scripts.
• National Academies of Sciences, Engineering, and Medicine. 2019. Reproducibility and Replicability in Science. Washington, DC: National Academies Press. DOI.
• Terashima, Seiji. 2026. “Entanglement Wedge Reconstruction without Holographic Quantum Error Correction.” arXiv:2607.08684v1, 9 July 2026. Versioned preprint.
• Wilkinson, Mark D., Michel Dumontier, IJsbrand Jan Aalbersberg, et al. 2016. “The FAIR Guiding Principles for Scientific Data Management and Stewardship.” Scientific Data 3, 160018. DOI.
• Witten, Edward. 1998. “Anti-de Sitter Space and Holography.” Advances in Theoretical and Mathematical Physics 2, 253–291. DOI; arXiv v2.
• Witten, Edward. 2026. “A Note on Corrections to Entanglement Wedge Reconstruction.” arXiv:2606.18639v3, 10 July 2026. Versioned preprint.