Add judge-readable DiffRes evidence logbook

Adds claim-level Markdown pages for the existing bounded stored-evidence package. No numerical evaluation is added.

MANIFEST.md

@@ -5,7 +5,9 @@
 This manifest records the provenance and SHA-256 identity of every copied
 public asset. The `evidence/` files and `UPSTREAM-MPL-2.0.txt` were copied
 byte-for-byte; wrapper documentation and the static page were authored for this
-package. No numerical evaluation was run while creating this package.
+package. The package also includes judge-readable Markdown logbook pages that
+restate the same evidence boundary and link to the preserved artefacts. No
+numerical evaluation was run while creating this package.
 
 ## Package-only verifier
 
@@ -52,7 +54,8 @@ this package:
 The package does not include the upstream source tree, any virtual environment
 or cache, numerical runners, model artefacts, partial GMS outputs, or any
 LGSSM, CIFAR-10, Lokta, timing, or other study materials. It contains no upload
-client or deployment script.
+client or deployment script. The `pages/` files are authored Markdown summaries
+of the included stored evidence, not additional numerical results.
 
 ## Integrity check
 

MANIFEST.sha256

@@ -1,9 +1,13 @@
-aa7e793d938e2371c7e759bd81e7bb8355336705dc92e28e1c7f89ddc89d971a  README.md
+8dbaa86b1ab3956c1a0cdbb3b2f7610118fe58a4256a4c3d506b5b8702f66b08  README.md
 c39f5f10df2a79c63f95eb000a60c29a6caf937372a127d4ed362125f693acd4  index.html
 9925f027238163d3b60fe3edbc991375b1b40da80556df3792d54844bb404038  LICENSE
 63dd00dfab08c4f4dd8ccd86cef24125b8aeefe3a8c28c30b9a00230afc9e12e  NOTICE
 3f3d9e0024b1921b067d6f7f88deb4a60cbe7a78e76c64e3f1d7fc3b779b9d04  UPSTREAM-MPL-2.0.txt
-75d21d15838ddc8627f7b2136adb08077e904979a06a0d3dc0970e997e4531e0  MANIFEST.md
+9e179806867991c80fdb7e399e9df28adaf09be77a2256eb1e012bdc51c3bf4e  MANIFEST.md
+d22a2938a186e21b861daf4ffe20a541ee096930c3110bfa0e407a1304270f36  pages/claim-1-gms-comparison/page.md
+37cb275e9744db4836365942d0adc15d33b6ae134fe94948908434df548b4ec1  pages/claim-2-consistency-theory/page.md
+f1410152219c521507a4820c81d83b370906cb27e8bf4d8756934617b789c0a0  pages/conclusion/page.md
+3607686204b9ce218929351aaf4abc8e59781397fe75d44da3e0f9d860ecceb3  pages/index.md
 50fa516668270ac13b332c7c539ce94cc45118faf775607051e94e4be86c5734  evidence/README.md
 c74851b88b5df723cbfb28edb48be09eb61e7a2dab086088f90d59591b68723b  evidence/gms-selected-ids-0-99.json
 8e9c7263415799f6f5cd456696290f5034ff8670c77e17951ea6be5f584c07ee  evidence/gms-selected-ids-0-99-validation.json

README.md

@@ -28,6 +28,11 @@ The files here are **historic stored outputs**, transferred without rerunning
 the numerical evaluation. This static site has no execution, upload, or model
 hosting component.
 
+The judge-readable, claim-level record is in the
+[evidence logbook](pages/index.md). It restates the concrete conditions,
+numbers, and limits below in Markdown because the challenge judge reads those
+pages rather than the static landing page or linked JSON files.
+
 ## What the evidence covers
 
 - Released implementation: [`zgbkdlm/diffres`](https://github.com/zgbkdlm/diffres)

pages/claim-1-gms-comparison/page.md

@@ -0,0 +1,31 @@
+# Claim 1 — benchmark outperformance
+
+## Challenge claim
+
+> Diffusion resampling method outperforms state-of-the-art differentiable resampling methods on multiple filtering and parameter estimation benchmarks.
+
+## Evidence examined
+
+The preserved record covers one paper-selected Gaussian-mixture configuration
+and compares Diffusion resampling with the multinomial baseline. It does not
+contain OT, Gumbel-Softmax, soft-resampling, filtering benchmarks, or parameter
+estimation benchmark suites.
+
+For stored Monte Carlo IDs `0`–`99`, the recorded sliced-Wasserstein L1 mean ±
+population SD was `0.0807686 ± 0.0212899` for Diffusion resampling and
+`0.0824334 ± 0.0250323` for multinomial. The stored squared posterior-mean
+residual L2 mean ± population SD was `0.0373908 ± 0.0299358` and
+`0.0378330 ± 0.0442877`, respectively. The paired 95% t intervals in the
+historic validation record cross zero for both metrics.
+
+## Record and limits
+
+The [raw 100-ID record](../../evidence/gms-selected-ids-0-99.json) and
+[post-hoc validation record](../../evidence/gms-selected-ids-0-99-validation.json)
+are preserved artefacts. The package verifier recomputes the raw record's JSON
+structure, configuration, ordered IDs, aggregates, paired aggregates, and
+lower-counts. It does not read the historical validation record, rerun DiffRes,
+or establish execution provenance.
+
+This single stored comparison does not establish the challenge claim's
+multi-benchmark, state-of-the-art, or statistical-significance statements.

pages/claim-2-consistency-theory/page.md

@@ -0,0 +1,17 @@
+# Claim 2 — consistent resampling distribution and proof
+
+## Challenge claim
+
+> Method provides a consistent resampling distribution with theoretical proof.
+
+## Evidence available in this logbook
+
+This package contains a stored finite Gaussian-mixture result and post-hoc
+arithmetic checks only. It does not include a theorem derivation, proof
+artefact, proof audit, source excerpt, or a numerical test that can establish
+distributional consistency.
+
+The stored Gaussian-mixture comparison therefore has no direct evidential
+bearing on the theoretical-proof claim. The
+[evidence guide](../../evidence/README.md) documents that the included verifier
+checks the stored record only and cannot establish paper-level claims.

pages/conclusion/page.md

@@ -0,0 +1,13 @@
+# Conclusion and evidence boundary
+
+The public artefacts preserve one 100-ID Gaussian-mixture comparison, its
+configuration, source pin, and post-hoc arithmetic checks. The numerical means
+recorded for Diffusion resampling are lower than the multinomial baseline for
+the two listed metrics, while the recorded paired intervals cross zero.
+
+The package is not a rerun, independent replication, complete Table 1 or Table
+2 reproduction, multi-baseline comparison, significance study, or proof audit.
+It contains no evidence for the theoretical-consistency claim and no basis for
+a broad outperformance conclusion. The preserved record, validator source, and
+self-contained verifier remain available for inspection through the artefacts
+linked from the index page.

pages/index.md

@@ -0,0 +1,33 @@
+# Diffusion differentiable resampling — evidence logbook
+
+**Paper.** [Diffusion differentiable resampling](https://openreview.net/forum?id=M0e5XORjAW), Jennifer R. Andersson and Zheng Zhao, [arXiv:2512.10401](https://arxiv.org/abs/2512.10401).
+
+**Claim source.** ICML 2026 challenge claims for `M0e5XORjAW`.
+
+**Evidence type.** Preserved historical numerical output plus post-hoc record validation. No numerical evaluation was rerun while assembling this Space.
+
+## Scope
+
+This logbook presents one stored Gaussian-mixture comparison from the released
+[`zgbkdlm/diffres`](https://github.com/zgbkdlm/diffres) implementation at
+commit `767effe3e755067eb8a04422597fbf37eb8ab754`. It contains Monte Carlo IDs
+`0`–`99`, 10,000 particles, 1,000 sliced-Wasserstein projections,
+probability-flow ODE `T=3`, 128 steps, and the Jentzen--Kloeden integrator. The
+comparison is Diffusion resampling versus the multinomial baseline only.
+
+## Evidence pages
+
+- [Claim 1: benchmark outperformance](claim-1-gms-comparison/page.md)
+- [Claim 2: consistency distribution and proof](claim-2-consistency-theory/page.md)
+- [Conclusion and evidence boundary](conclusion/page.md)
+
+## Inspectable artefacts
+
+- [Stored 100-ID record](../evidence/gms-selected-ids-0-99.json)
+- [Historic validation record](../evidence/gms-selected-ids-0-99-validation.json)
+- [Self-contained record verifier](../evidence/verify_stored_gms_evidence.py)
+- [Evidence guide](../evidence/README.md)
+
+The pages state the available evidence and its limits directly. Linked files
+allow independent inspection but do not substitute for a numerical rerun,
+complete benchmark comparison, significance study, or proof audit.