Publish bounded DiffRes stored-evidence package

Static evidence package for one stored 100-ID Gaussian-mixture comparison. It verifies the stored record only and explicitly does not claim a source rerun, paper-level reproduction, significance, or broad outperformance.

LICENSE

@@ -0,0 +1,22 @@
+MIT License
+
+Copyright (c) 2026 DiffRes publication package contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so,
+subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

MANIFEST.md

@@ -0,0 +1,66 @@
+# Source and artefact manifest
+
+## Package purpose
+
+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-only verifier
+
+[`evidence/verify_stored_gms_evidence.py`](evidence/verify_stored_gms_evidence.py)
+and its focused standard-library test
+[`evidence/test_verify_stored_gms_evidence.py`](evidence/test_verify_stored_gms_evidence.py)
+are package-authored. They recompute the immutable raw record's JSON structure,
+configuration, ordered IDs, descriptive aggregates, paired aggregates, and
+lower-counts without a source checkout, SciPy, network access, or numerical
+evaluation. The verifier itself has no output-file dependency; its focused test
+uses only temporary local fixtures. They do not validate the recorded upstream
+hashes against source files, prove provenance, or establish paper-level claims.
+
+## Copied public assets
+
+| Package path | Source role | SHA-256 | Transfer |
+| --- | --- | --- | --- |
+| `evidence/gms-selected-ids-0-99.json` | Historical 100-ID GMS raw result | `c74851b88b5df723cbfb28edb48be09eb61e7a2dab086088f90d59591b68723b` | Byte-for-byte copy |
+| `evidence/gms-selected-ids-0-99-validation.json` | Historical post-hoc validation record | `8e9c7263415799f6f5cd456696290f5034ff8670c77e17951ea6be5f584c07ee` | Byte-for-byte copy |
+| `evidence/validate_gms_selected_result.py` | Separate post-hoc validator source | `80841d7e4395f60bcbf06880a92288ad7bfdd7e022ea60876911277f3b3e69e5` | Byte-for-byte copy |
+| `UPSTREAM-MPL-2.0.txt` | Upstream MPL-2.0 licence text | `3f3d9e0024b1921b067d6f7f88deb4a60cbe7a78e76c64e3f1d7fc3b779b9d04` | Byte-for-byte copy |
+
+## Recorded released-source identity (not vendored)
+
+The raw result and validation record identify the upstream repository as
+`https://github.com/zgbkdlm/diffres.git`, commit
+`767effe3e755067eb8a04422597fbf37eb8ab754`. The validator's historical source
+hash checks cover the following released files; they are **not** copied into
+this package:
+
+| Released path | SHA-256 |
+| --- | --- |
+| `diffres/integators.py` | `4cbcdba683cf27532b21acfa9c3295354c52cc2f4494b4063a0459629eb72954` |
+| `diffres/resampling.py` | `12b33fe23e1f62e18909cbcff9fa45db36ed65f6f943668728c9fe61795aa06e` |
+| `diffres/tools.py` | `54c60e1ac4f484532f51be57a63aacbdfb7a53ef8cef621495176ea7a843a5df` |
+| `experiments/gms/baselines.py` | `7b258df0ab4a7ca8eb5c91f6d5efbabbe359e0130db1cb9e4d3263a9241f1d5c` |
+| `experiments/gms/diffusion.py` | `886496bab31d254ac8030286ea350e48db817bd680cccc8cdda50c58a94ba774` |
+| `experiments/rnd_keys.npy` | `6e60685445936e0380220f34fc3fd15b5b7720070dc8c8f7f1148575de4ff5df` |
+| `experiments/run_gms.sh` | `c01a4c88a3ef05d6a5ffb585b9d4233b1b1a9e8c1c46d40360eab44963b3eee5` |
+| `experiments/summary/print_gms_errs.py` | `1df81cadaaf35ca5b26e8bd1ce6fc19aa4a60a2f7e161075087e518b3a0d17cd` |
+
+## Deliberate exclusions
+
+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.
+
+## Integrity check
+
+From the package root, run:
+
+```sh
+shasum -a 256 -c MANIFEST.sha256
+```
+
+`MANIFEST.sha256` covers every package file except itself. It is a packaging
+integrity check only; it does not validate the underlying numerical result.

MANIFEST.sha256

@@ -0,0 +1,12 @@
+aa7e793d938e2371c7e759bd81e7bb8355336705dc92e28e1c7f89ddc89d971a  README.md
+c39f5f10df2a79c63f95eb000a60c29a6caf937372a127d4ed362125f693acd4  index.html
+9925f027238163d3b60fe3edbc991375b1b40da80556df3792d54844bb404038  LICENSE
+63dd00dfab08c4f4dd8ccd86cef24125b8aeefe3a8c28c30b9a00230afc9e12e  NOTICE
+3f3d9e0024b1921b067d6f7f88deb4a60cbe7a78e76c64e3f1d7fc3b779b9d04  UPSTREAM-MPL-2.0.txt
+75d21d15838ddc8627f7b2136adb08077e904979a06a0d3dc0970e997e4531e0  MANIFEST.md
+50fa516668270ac13b332c7c539ce94cc45118faf775607051e94e4be86c5734  evidence/README.md
+c74851b88b5df723cbfb28edb48be09eb61e7a2dab086088f90d59591b68723b  evidence/gms-selected-ids-0-99.json
+8e9c7263415799f6f5cd456696290f5034ff8670c77e17951ea6be5f584c07ee  evidence/gms-selected-ids-0-99-validation.json
+80841d7e4395f60bcbf06880a92288ad7bfdd7e022ea60876911277f3b3e69e5  evidence/validate_gms_selected_result.py
+7359767586cc5f5410b940de359adac14b6ff69acc2570ab8b7b0d3e55fd9a49  evidence/test_verify_stored_gms_evidence.py
+e030156293c8bf77f211f31220ca4fcd96c9f8850267f32cd315b5551b6931d6  evidence/verify_stored_gms_evidence.py

NOTICE

@@ -0,0 +1,29 @@
+# Notice and attribution
+
+## Package scope
+
+This repository is a local static publication package. Its wrapper-authored
+documentation and presentation material are licensed under the MIT Licence in
+[`LICENSE`](LICENSE). The evidence JSON records and validator are preserved as
+historic study artefacts; this notice does not grant rights that their original
+contributors do not hold.
+
+## Upstream implementation
+
+The study evidence refers to the released implementation:
+
+- Repository: <https://github.com/zgbkdlm/diffres>
+- Revision: `767effe3e755067eb8a04422597fbf37eb8ab754`
+- Licence: Mozilla Public License 2.0 (MPL-2.0)
+
+No upstream implementation source files are redistributed in this package.
+[`UPSTREAM-MPL-2.0.txt`](UPSTREAM-MPL-2.0.txt) is a byte-for-byte copy of the
+upstream licence text, included for attribution and reference only.
+
+## Paper
+
+Jennifer Rosina Andersson and Zheng Zhao, *Diffusion differentiable
+resampling*, arXiv:2512.10401, OpenReview:M0e5XORjAW.
+
+This package is an evidence wrapper, not an official paper artefact or an
+endorsement by the paper authors.

README.md

@@ -1,10 +1,96 @@
 ---
-title: Repro Diffusion Differentiable Resampling
-emoji: ⚡
-colorFrom: indigo
+title: "DiffRes: traceable 100-ID Gaussian-mixture study"
+emoji: "🔬"
+colorFrom: blue
 colorTo: indigo
 sdk: static
 pinned: false
+tags:
+  - icml2026-repro
+  - paper-M0e5XORjAW
+  - reproducibility
+  - diffusion
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# DiffRes: traceable 100-ID Gaussian-mixture study
+
+This is a static, local publication package for **Diffusion differentiable
+resampling** by Jennifer Rosina Andersson and Zheng Zhao
+([arXiv:2512.10401](https://arxiv.org/abs/2512.10401),
+[OpenReview:M0e5XORjAW](https://openreview.net/forum?id=M0e5XORjAW)).
+
+**Evidence statement.** A local study wrapper using released implementation,
+stored numerical random keys and pinned source config recorded a traceable 100-ID
+Gaussian-mixture result; a separate validator recomputed raw-record aggregates,
+paired intervals, and source hashes.
+
+The files here are **historic stored outputs**, transferred without rerunning
+the numerical evaluation. This static site has no execution, upload, or model
+hosting component.
+
+## What the evidence covers
+
+- Released implementation: [`zgbkdlm/diffres`](https://github.com/zgbkdlm/diffres)
+  at commit `767effe3e755067eb8a04422597fbf37eb8ab754` (MPL-2.0).
+- One paper-selected Gaussian-mixture configuration, with stored 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.
+- A paired comparison between diffusion resampling and the multinomial
+  baseline only.
+- A post-hoc, standard-Python validator that checks strict JSON, the 100 raw
+  records, reported aggregates, paired intervals, the recorded source pin, and
+  hashes of the released source files.
+
+| Historical stored aggregate (lower is better) | Diffusion resampling | Multinomial baseline |
+| --- | ---: | ---: |
+| Sliced Wasserstein L1, mean ± population SD | 0.0807686 ± 0.0212899 | 0.0824334 ± 0.0250323 |
+| Squared posterior-mean residual L2, mean ± population SD | 0.0373908 ± 0.0299358 | 0.0378330 ± 0.0442877 |
+
+The stored paired 95% t intervals cross zero for both metrics. They are not
+evidence of a statistically significant improvement.
+
+## Evidence files
+
+- [Raw 100-ID result](evidence/gms-selected-ids-0-99.json) — preserved
+  per-ID values, source provenance, configuration, and recorded aggregates.
+- [Post-hoc validation record](evidence/gms-selected-ids-0-99-validation.json)
+  — the historic validator output.
+- [Validator source](evidence/validate_gms_selected_result.py) — copied
+  byte-for-byte as an audit artefact; it does not rerun the JAX evaluation.
+- [Self-contained stored-evidence verifier](evidence/verify_stored_gms_evidence.py)
+  — a standard-library-only check of the JSON structure and recorded
+  arithmetic; run it with `python3 evidence/verify_stored_gms_evidence.py`.
+- [Evidence guide](evidence/README.md) — lineage, portability, and limits.
+- [Source and artefact manifest](MANIFEST.md) and
+  [package SHA-256 manifest](MANIFEST.sha256).
+
+## Limits
+
+This is **not** an exact reproduction, an independent replication, a complete
+Table 1 or Table 2 reproduction, a publication score, or a broad
+outperformance claim. It does not evaluate the OT, Gumbel-Softmax, or
+soft-resampling baselines, and it does not rerun the numerical study in this
+package. The included validator recomputes historic raw-record statistics and
+checks the recorded source identity; it is not a second numerical run. The
+self-contained verifier recomputes the stored record's configuration, ordered
+IDs, aggregates, and lower-counts without a source checkout or SciPy. It does
+not validate source hashes, execution provenance, paper-level results, or any
+comparative claim.
+
+## Safety and selection note
+
+The selected material is a benign numerical-methods study package: static
+HTML/Markdown, two JSON evidence records, a non-networked validator, and
+licensing/provenance files. The transfer excludes the upstream source tree,
+virtual environments, caches, runners, model artefacts, partial studies, and
+all LGSSM material. A pre-transfer content scan found no private local paths,
+credential-shaped values, upload clients, or source files from excluded study
+areas. The validator's only generic token-pattern match is a JSON-parser parameter,
+not credential handling.
+
+## Licence and attribution
+
+The package-authored wrapper material is under the [MIT Licence](LICENSE).
+The upstream DiffRes implementation is not redistributed here; it remains
+available under MPL-2.0, with attribution in [NOTICE](NOTICE) and a copy of
+the upstream licence at [UPSTREAM-MPL-2.0.txt](UPSTREAM-MPL-2.0.txt).

UPSTREAM-MPL-2.0.txt

@@ -0,0 +1,373 @@
+Mozilla Public License Version 2.0
+==================================
+
+1. Definitions
+--------------
+
+1.1. "Contributor"
+    means each individual or legal entity that creates, contributes to
+    the creation of, or owns Covered Software.
+
+1.2. "Contributor Version"
+    means the combination of the Contributions of others (if any) used
+    by a Contributor and that particular Contributor's Contribution.
+
+1.3. "Contribution"
+    means Covered Software of a particular Contributor.
+
+1.4. "Covered Software"
+    means Source Code Form to which the initial Contributor has attached
+    the notice in Exhibit A, the Executable Form of such Source Code
+    Form, and Modifications of such Source Code Form, in each case
+    including portions thereof.
+
+1.5. "Incompatible With Secondary Licenses"
+    means
+
+    (a) that the initial Contributor has attached the notice described
+        in Exhibit B to the Covered Software; or
+
+    (b) that the Covered Software was made available under the terms of
+        version 1.1 or earlier of the License, but not also under the
+        terms of a Secondary License.
+
+1.6. "Executable Form"
+    means any form of the work other than Source Code Form.
+
+1.7. "Larger Work"
+    means a work that combines Covered Software with other material, in
+    a separate file or files, that is not Covered Software.
+
+1.8. "License"
+    means this document.
+
+1.9. "Licensable"
+    means having the right to grant, to the maximum extent possible,
+    whether at the time of the initial grant or subsequently, any and
+    all of the rights conveyed by this License.
+
+1.10. "Modifications"
+    means any of the following:
+
+    (a) any file in Source Code Form that results from an addition to,
+        deletion from, or modification of the contents of Covered
+        Software; or
+
+    (b) any new file in Source Code Form that contains any Covered
+        Software.
+
+1.11. "Patent Claims" of a Contributor
+    means any patent claim(s), including without limitation, method,
+    process, and apparatus claims, in any patent Licensable by such
+    Contributor that would be infringed, but for the grant of the
+    License, by the making, using, selling, offering for sale, having
+    made, import, or transfer of either its Contributions or its
+    Contributor Version.
+
+1.12. "Secondary License"
+    means either the GNU General Public License, Version 2.0, the GNU
+    Lesser General Public License, Version 2.1, the GNU Affero General
+    Public License, Version 3.0, or any later versions of those
+    licenses.
+
+1.13. "Source Code Form"
+    means the form of the work preferred for making modifications.
+
+1.14. "You" (or "Your")
+    means an individual or a legal entity exercising rights under this
+    License. For legal entities, "You" includes any entity that
+    controls, is controlled by, or is under common control with You. For
+    purposes of this definition, "control" means (a) the power, direct
+    or indirect, to cause the direction or management of such entity,
+    whether by contract or otherwise, or (b) ownership of more than
+    fifty percent (50%) of the outstanding shares or beneficial
+    ownership of such entity.
+
+2. License Grants and Conditions
+--------------------------------
+
+2.1. Grants
+
+Each Contributor hereby grants You a world-wide, royalty-free,
+non-exclusive license:
+
+(a) under intellectual property rights (other than patent or trademark)
+    Licensable by such Contributor to use, reproduce, make available,
+    modify, display, perform, distribute, and otherwise exploit its
+    Contributions, either on an unmodified basis, with Modifications, or
+    as part of a Larger Work; and
+
+(b) under Patent Claims of such Contributor to make, use, sell, offer
+    for sale, have made, import, and otherwise transfer either its
+    Contributions or its Contributor Version.
+
+2.2. Effective Date
+
+The licenses granted in Section 2.1 with respect to any Contribution
+become effective for each Contribution on the date the Contributor first
+distributes such Contribution.
+
+2.3. Limitations on Grant Scope
+
+The licenses granted in this Section 2 are the only rights granted under
+this License. No additional rights or licenses will be implied from the
+distribution or licensing of Covered Software under this License.
+Notwithstanding Section 2.1(b) above, no patent license is granted by a
+Contributor:
+
+(a) for any code that a Contributor has removed from Covered Software;
+    or
+
+(b) for infringements caused by: (i) Your and any other third party's
+    modifications of Covered Software, or (ii) the combination of its
+    Contributions with other software (except as part of its Contributor
+    Version); or
+
+(c) under Patent Claims infringed by Covered Software in the absence of
+    its Contributions.
+
+This License does not grant any rights in the trademarks, service marks,
+or logos of any Contributor (except as may be necessary to comply with
+the notice requirements in Section 3.4).
+
+2.4. Subsequent Licenses
+
+No Contributor makes additional grants as a result of Your choice to
+distribute the Covered Software under a subsequent version of this
+License (see Section 10.2) or under the terms of a Secondary License (if
+permitted under the terms of Section 3.3).
+
+2.5. Representation
+
+Each Contributor represents that the Contributor believes its
+Contributions are its original creation(s) or it has sufficient rights
+to grant the rights to its Contributions conveyed by this License.
+
+2.6. Fair Use
+
+This License is not intended to limit any rights You have under
+applicable copyright doctrines of fair use, fair dealing, or other
+equivalents.
+
+2.7. Conditions
+
+Sections 3.1, 3.2, 3.3, and 3.4 are conditions of the licenses granted
+in Section 2.1.
+
+3. Responsibilities
+-------------------
+
+3.1. Distribution of Source Form
+
+All distribution of Covered Software in Source Code Form, including any
+Modifications that You create or to which You contribute, must be under
+the terms of this License. You must inform recipients that the Source
+Code Form of the Covered Software is governed by the terms of this
+License, and how they can obtain a copy of this License. You may not
+attempt to alter or restrict the recipients' rights in the Source Code
+Form.
+
+3.2. Distribution of Executable Form
+
+If You distribute Covered Software in Executable Form then:
+
+(a) such Covered Software must also be made available in Source Code
+    Form, as described in Section 3.1, and You must inform recipients of
+    the Executable Form how they can obtain a copy of such Source Code
+    Form by reasonable means in a timely manner, at a charge no more
+    than the cost of distribution to the recipient; and
+
+(b) You may distribute such Executable Form under the terms of this
+    License, or sublicense it under different terms, provided that the
+    license for the Executable Form does not attempt to limit or alter
+    the recipients' rights in the Source Code Form under this License.
+
+3.3. Distribution of a Larger Work
+
+You may create and distribute a Larger Work under terms of Your choice,
+provided that You also comply with the requirements of this License for
+the Covered Software. If the Larger Work is a combination of Covered
+Software with a work governed by one or more Secondary Licenses, and the
+Covered Software is not Incompatible With Secondary Licenses, this
+License permits You to additionally distribute such Covered Software
+under the terms of such Secondary License(s), so that the recipient of
+the Larger Work may, at their option, further distribute the Covered
+Software under the terms of either this License or such Secondary
+License(s).
+
+3.4. Notices
+
+You may not remove or alter the substance of any license notices
+(including copyright notices, patent notices, disclaimers of warranty,
+or limitations of liability) contained within the Source Code Form of
+the Covered Software, except that You may alter any license notices to
+the extent required to remedy known factual inaccuracies.
+
+3.5. Application of Additional Terms
+
+You may choose to offer, and to charge a fee for, warranty, support,
+indemnity or liability obligations to one or more recipients of Covered
+Software. However, You may do so only on Your own behalf, and not on
+behalf of any Contributor. You must make it absolutely clear that any
+such warranty, support, indemnity, or liability obligation is offered by
+You alone, and You hereby agree to indemnify every Contributor for any
+liability incurred by such Contributor as a result of warranty, support,
+indemnity or liability terms You offer. You may include additional
+disclaimers of warranty and limitations of liability specific to any
+jurisdiction.
+
+4. Inability to Comply Due to Statute or Regulation
+---------------------------------------------------
+
+If it is impossible for You to comply with any of the terms of this
+License with respect to some or all of the Covered Software due to
+statute, judicial order, or regulation then You must: (a) comply with
+the terms of this License to the maximum extent possible; and (b)
+describe the limitations and the code they affect. Such description must
+be placed in a text file included with all distributions of the Covered
+Software under this License. Except to the extent prohibited by statute
+or regulation, such description must be sufficiently detailed for a
+recipient of ordinary skill to be able to understand it.
+
+5. Termination
+--------------
+
+5.1. The rights granted under this License will terminate automatically
+if You fail to comply with any of its terms. However, if You become
+compliant, then the rights granted under this License from a particular
+Contributor are reinstated (a) provisionally, unless and until such
+Contributor explicitly and finally terminates Your grants, and (b) on an
+ongoing basis, if such Contributor fails to notify You of the
+non-compliance by some reasonable means prior to 60 days after You have
+come back into compliance. Moreover, Your grants from a particular
+Contributor are reinstated on an ongoing basis if such Contributor
+notifies You of the non-compliance by some reasonable means, this is the
+first time You have received notice of non-compliance with this License
+from such Contributor, and You become compliant prior to 30 days after
+Your receipt of the notice.
+
+5.2. If You initiate litigation against any entity by asserting a patent
+infringement claim (excluding declaratory judgment actions,
+counter-claims, and cross-claims) alleging that a Contributor Version
+directly or indirectly infringes any patent, then the rights granted to
+You by any and all Contributors for the Covered Software under Section
+2.1 of this License shall terminate.
+
+5.3. In the event of termination under Sections 5.1 or 5.2 above, all
+end user license agreements (excluding distributors and resellers) which
+have been validly granted by You or Your distributors under this License
+prior to termination shall survive termination.
+
+************************************************************************
+*                                                                      *
+*  6. Disclaimer of Warranty                                           *
+*  -------------------------                                           *
+*                                                                      *
+*  Covered Software is provided under this License on an "as is"       *
+*  basis, without warranty of any kind, either expressed, implied, or  *
+*  statutory, including, without limitation, warranties that the       *
+*  Covered Software is free of defects, merchantable, fit for a        *
+*  particular purpose or non-infringing. The entire risk as to the     *
+*  quality and performance of the Covered Software is with You.        *
+*  Should any Covered Software prove defective in any respect, You     *
+*  (not any Contributor) assume the cost of any necessary servicing,   *
+*  repair, or correction. This disclaimer of warranty constitutes an   *
+*  essential part of this License. No use of any Covered Software is   *
+*  authorized under this License except under this disclaimer.         *
+*                                                                      *
+************************************************************************
+
+************************************************************************
+*                                                                      *
+*  7. Limitation of Liability                                          *
+*  --------------------------                                          *
+*                                                                      *
+*  Under no circumstances and under no legal theory, whether tort      *
+*  (including negligence), contract, or otherwise, shall any           *
+*  Contributor, or anyone who distributes Covered Software as          *
+*  permitted above, be liable to You for any direct, indirect,         *
+*  special, incidental, or consequential damages of any character      *
+*  including, without limitation, damages for lost profits, loss of    *
+*  goodwill, work stoppage, computer failure or malfunction, or any    *
+*  and all other commercial damages or losses, even if such party      *
+*  shall have been informed of the possibility of such damages. This   *
+*  limitation of liability shall not apply to liability for death or   *
+*  personal injury resulting from such party's negligence to the       *
+*  extent applicable law prohibits such limitation. Some               *
+*  jurisdictions do not allow the exclusion or limitation of           *
+*  incidental or consequential damages, so this exclusion and          *
+*  limitation may not apply to You.                                    *
+*                                                                      *
+************************************************************************
+
+8. Litigation
+-------------
+
+Any litigation relating to this License may be brought only in the
+courts of a jurisdiction where the defendant maintains its principal
+place of business and such litigation shall be governed by laws of that
+jurisdiction, without reference to its conflict-of-law provisions.
+Nothing in this Section shall prevent a party's ability to bring
+cross-claims or counter-claims.
+
+9. Miscellaneous
+----------------
+
+This License represents the complete agreement concerning the subject
+matter hereof. If any provision of this License is held to be
+unenforceable, such provision shall be reformed only to the extent
+necessary to make it enforceable. Any law or regulation which provides
+that the language of a contract shall be construed against the drafter
+shall not be used to construe this License against a Contributor.
+
+10. Versions of the License
+---------------------------
+
+10.1. New Versions
+
+Mozilla Foundation is the license steward. Except as provided in Section
+10.3, no one other than the license steward has the right to modify or
+publish new versions of this License. Each version will be given a
+distinguishing version number.
+
+10.2. Effect of New Versions
+
+You may distribute the Covered Software under the terms of the version
+of the License under which You originally received the Covered Software,
+or under the terms of any subsequent version published by the license
+steward.
+
+10.3. Modified Versions
+
+If you create software not governed by this License, and you want to
+create a new license for such software, you may create and use a
+modified version of this License if you rename the license and remove
+any references to the name of the license steward (except to note that
+such modified license differs from this License).
+
+10.4. Distributing Source Code Form that is Incompatible With Secondary
+Licenses
+
+If You choose to distribute Source Code Form that is Incompatible With
+Secondary Licenses under the terms of this version of the License, the
+notice described in Exhibit B of this License must be attached.
+
+Exhibit A - Source Code Form License Notice
+-------------------------------------------
+
+  This Source Code Form is subject to the terms of the Mozilla Public
+  License, v. 2.0. If a copy of the MPL was not distributed with this
+  file, You can obtain one at https://mozilla.org/MPL/2.0/.
+
+If it is not possible or desirable to put the notice in a particular
+file, then You may include the notice in a location (such as a LICENSE
+file in a relevant directory) where a recipient would be likely to look
+for such a notice.
+
+You may add additional accurate notices of copyright ownership.
+
+Exhibit B - "Incompatible With Secondary Licenses" Notice
+---------------------------------------------------------
+
+  This Source Code Form is "Incompatible With Secondary Licenses", as
+  defined by the Mozilla Public License, v. 2.0.

evidence/README.md

@@ -0,0 +1,64 @@
+# Evidence guide
+
+## Lineage
+
+The public evidence set is intentionally small:
+
+1. [`gms-selected-ids-0-99.json`](gms-selected-ids-0-99.json) is the
+   historical stored result: 100 ordered Gaussian-mixture records, the selected
+   configuration, source provenance, and reported aggregates.
+2. [`validate_gms_selected_result.py`](validate_gms_selected_result.py) is the
+   separate post-hoc validator, copied byte-for-byte from the local study.
+3. [`gms-selected-ids-0-99-validation.json`](gms-selected-ids-0-99-validation.json)
+   is the historical output from that validator.
+4. [`verify_stored_gms_evidence.py`](verify_stored_gms_evidence.py) is a new,
+   standard-library-only verifier for the preserved JSON record.
+5. [`test_verify_stored_gms_evidence.py`](test_verify_stored_gms_evidence.py)
+   is its focused standard-library test suite.
+
+The validator is intentionally separate from the JAX numerical runner. It
+recomputes strict-JSON checks, the 100 ordered raw records, aggregate statistics,
+paired intervals, and the recorded hashes of released source files. It does not
+run the numerical evaluation.
+
+## Portability
+
+The validator is preserved as an audit artefact, not as a standalone Space
+application. It expects the released DiffRes checkout layout at the recorded
+commit and checks its Git identity and eight source/input hashes. The upstream
+source is deliberately not vendored in this package. Consequently, the bundled
+historical validation JSON is the inspection record for this static package;
+running the validator requires a separately obtained clean checkout at the
+recorded pin and a Python environment with SciPy.
+
+The self-contained verifier fills the package-only gap without changing the
+historical artefacts. From the package root, run:
+
+```sh
+python3 evidence/verify_stored_gms_evidence.py
+python3 evidence/test_verify_stored_gms_evidence.py
+```
+
+It strictly parses the raw JSON, rejects duplicate keys and non-finite values,
+checks the declared configuration and ordered IDs `0`–`99`, and recomputes the
+method/metric aggregates, paired aggregates, and lower-counts. It has no
+network, source-checkout, third-party-package, numerical-run, or output-file
+dependency. It cannot validate the claimed source hashes against a checkout,
+prove execution provenance, or establish a paper result, statistical
+significance, or outperformance claim.
+
+## Interpretation limits
+
+The stored result covers the selected 100-ID Gaussian-mixture configuration and
+the multinomial baseline. It does not supply a full Table 1 or Table 2 result,
+an independent numerical rerun, other baseline comparisons, a significance
+claim, or an outperformance claim. The paired 95% t intervals recorded by the
+validator cross zero for both metrics.
+
+## Asset integrity
+
+Each transferred asset is listed with its source path, SHA-256, and transfer
+mode in the repository-root [manifest](../MANIFEST.md). The package-level
+[SHA-256 manifest](../MANIFEST.sha256) enables a local integrity check without
+executing either validator. The focused verifier tests run only with the Python
+standard library.

evidence/gms-selected-ids-0-99-validation.json

@@ -0,0 +1,112 @@
+{
+  "checked_at_utc": "2026-07-19T11:19:22.402493Z",
+  "checks": {
+    "current_source_identity_and_hashes": {
+      "checked_author_source_sha256": {
+        "diffres/integators.py": "4cbcdba683cf27532b21acfa9c3295354c52cc2f4494b4063a0459629eb72954",
+        "diffres/resampling.py": "12b33fe23e1f62e18909cbcff9fa45db36ed65f6f943668728c9fe61795aa06e",
+        "diffres/tools.py": "54c60e1ac4f484532f51be57a63aacbdfb7a53ef8cef621495176ea7a843a5df",
+        "experiments/gms/baselines.py": "7b258df0ab4a7ca8eb5c91f6d5efbabbe359e0130db1cb9e4d3263a9241f1d5c",
+        "experiments/gms/diffusion.py": "886496bab31d254ac8030286ea350e48db817bd680cccc8cdda50c58a94ba774",
+        "experiments/rnd_keys.npy": "6e60685445936e0380220f34fc3fd15b5b7720070dc8c8f7f1148575de4ff5df",
+        "experiments/run_gms.sh": "c01a4c88a3ef05d6a5ffb585b9d4233b1b1a9e8c1c46d40360eab44963b3eee5",
+        "experiments/summary/print_gms_errs.py": "1df81cadaaf35ca5b26e8bd1ce6fc19aa4a60a2f7e161075087e518b3a0d17cd"
+      },
+      "current_commit": "767effe3e755067eb8a04422597fbf37eb8ab754",
+      "current_origin_matches_expected": true,
+      "current_tracked_worktree_clean": true
+    },
+    "raw_record_integrity_and_aggregate_recomputation": {
+      "diffusion_lower_is_better_wins": {
+        "posterior_mean_residual_squared_l2": 51,
+        "sliced_wasserstein_l1": 54
+      },
+      "paired_diffusion_minus_multinomial_95_percent_t_intervals": {
+        "posterior_mean_residual_squared_l2": {
+          "critical_value": 1.9842169515086827,
+          "lower": -0.007123449582480549,
+          "mean": -0.00044225546225615345,
+          "sample_count": 100,
+          "sample_standard_deviation": 0.03367169157155121,
+          "standard_error": 0.003367169157155121,
+          "upper": 0.006238938657968242
+        },
+        "sliced_wasserstein_l1": {
+          "critical_value": 1.9842169515086827,
+          "lower": -0.004659334515850839,
+          "mean": -0.0016648057831093993,
+          "sample_count": 100,
+          "sample_standard_deviation": 0.015091740499770325,
+          "standard_error": 0.0015091740499770326,
+          "upper": 0.0013297229496320405
+        }
+      },
+      "recomputed_aggregate": {
+        "diffusion_resampling": {
+          "posterior_mean_residual_squared_l2": {
+            "maximum": 0.13674925583503703,
+            "mean": 0.037390775125730545,
+            "minimum": 0.003946685433082616,
+            "population_standard_deviation": 0.029935755902808506
+          },
+          "sliced_wasserstein_l1": {
+            "maximum": 0.13845705348834103,
+            "mean": 0.08076861937013251,
+            "minimum": 0.049275986886894586,
+            "population_standard_deviation": 0.021289921970778995
+          }
+        },
+        "multinomial_baseline": {
+          "posterior_mean_residual_squared_l2": {
+            "maximum": 0.34224144904965786,
+            "mean": 0.0378330305879867,
+            "minimum": 0.00328472168500068,
+            "population_standard_deviation": 0.044287731038823226
+          },
+          "sliced_wasserstein_l1": {
+            "maximum": 0.19381190521839362,
+            "mean": 0.08243342515324191,
+            "minimum": 0.04778862503625366,
+            "population_standard_deviation": 0.02503231412513419
+          }
+        }
+      },
+      "recomputed_paired_aggregate": {
+        "posterior_mean_residual_squared_l2": {
+          "maximum": 0.10671040850944102,
+          "mean": -0.00044225546225615345,
+          "minimum": -0.24669590547035306,
+          "population_standard_deviation": 0.03350291009982232
+        },
+        "sliced_wasserstein_l1": {
+          "maximum": 0.041779128105039046,
+          "mean": -0.0016648057831093993,
+          "minimum": -0.07544369540487143,
+          "population_standard_deviation": 0.015016092201344656
+        }
+      },
+      "record_count": 100
+    },
+    "selected_protocol": "passed",
+    "strict_json": "passed"
+  },
+  "claim_boundary": "This check independently recomputes the recorded aggregates and paired intervals from the 100 raw JSON records while checking the current pinned source identity. It does not rerun the numerical experiment, independently reproduce the paper, or establish an outperformance or significance claim.",
+  "input": {
+    "path": "study/gms-selected-ids-0-99.json",
+    "sha256": "c74851b88b5df723cbfb28edb48be09eb61e7a2dab086088f90d59591b68723b"
+  },
+  "schema_version": 1,
+  "validation_status": "passed",
+  "validation_type": "post-hoc integrity and aggregate recomputation; no numerical rerun",
+  "validator": {
+    "path": "study/validate_gms_selected_result.py",
+    "platform": {
+      "machine": "arm64",
+      "release": "25.5.0",
+      "system": "Darwin"
+    },
+    "python_version": "3.13.14 (main, Jun 23 2026, 15:20:43) [Clang 22.1.3 ]",
+    "scipy_version": "1.16.2",
+    "sha256": "80841d7e4395f60bcbf06880a92288ad7bfdd7e022ea60876911277f3b3e69e5"
+  }
+}

evidence/gms-selected-ids-0-99.json

@@ -0,0 +1,1319 @@
+{
+  "aggregate": {
+    "diffusion_lower_is_better_wins": {
+      "posterior_mean_residual_squared_l2": 51,
+      "sliced_wasserstein_l1": 54
+    },
+    "diffusion_resampling": {
+      "posterior_mean_residual_squared_l2": {
+        "maximum": 0.13674925583503703,
+        "mean": 0.037390775125730545,
+        "minimum": 0.003946685433082616,
+        "population_standard_deviation": 0.029935755902808506
+      },
+      "sliced_wasserstein_l1": {
+        "maximum": 0.13845705348834103,
+        "mean": 0.08076861937013251,
+        "minimum": 0.049275986886894586,
+        "population_standard_deviation": 0.021289921970778995
+      }
+    },
+    "multinomial_baseline": {
+      "posterior_mean_residual_squared_l2": {
+        "maximum": 0.34224144904965786,
+        "mean": 0.037833030587986684,
+        "minimum": 0.00328472168500068,
+        "population_standard_deviation": 0.04428773103882322
+      },
+      "sliced_wasserstein_l1": {
+        "maximum": 0.19381190521839362,
+        "mean": 0.08243342515324191,
+        "minimum": 0.04778862503625366,
+        "population_standard_deviation": 0.025032314125134195
+      }
+    },
+    "paired_diffusion_minus_multinomial": {
+      "posterior_mean_residual_squared_l2": {
+        "maximum": 0.10671040850944102,
+        "mean": -0.0004422554622561538,
+        "minimum": -0.24669590547035306,
+        "population_standard_deviation": 0.03350291009982232
+      },
+      "sliced_wasserstein_l1": {
+        "maximum": 0.041779128105039046,
+        "mean": -0.0016648057831093993,
+        "minimum": -0.07544369540487143,
+        "population_standard_deviation": 0.015016092201344656
+      }
+    }
+  },
+  "claim_boundary": "Uses the fixed Jentzen--Kloeden ODE grid selection and 100 of the author's 100 stored Monte Carlo IDs. Its particle count and SWD projections match the paper-scale source grid. It is source-faithful and reproduction-compatible with the selected 100-ID source protocol because it uses exactly author IDs 0\u201399 at paper-scale inputs. It does not by itself establish an outperformance claim.",
+  "configuration": {
+    "author_keys_path": "experiments/rnd_keys.npy",
+    "components": 5,
+    "diffusion_a": -1.0,
+    "diffusion_steps": 128,
+    "dimension": 8,
+    "execution": "one CPU process; compiled functions shared across ids",
+    "integrator": "jentzen_and_kloeden",
+    "mc_ids": [
+      0,
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      7,
+      8,
+      9,
+      10,
+      11,
+      12,
+      13,
+      14,
+      15,
+      16,
+      17,
+      18,
+      19,
+      20,
+      21,
+      22,
+      23,
+      24,
+      25,
+      26,
+      27,
+      28,
+      29,
+      30,
+      31,
+      32,
+      33,
+      34,
+      35,
+      36,
+      37,
+      38,
+      39,
+      40,
+      41,
+      42,
+      43,
+      44,
+      45,
+      46,
+      47,
+      48,
+      49,
+      50,
+      51,
+      52,
+      53,
+      54,
+      55,
+      56,
+      57,
+      58,
+      59,
+      60,
+      61,
+      62,
+      63,
+      64,
+      65,
+      66,
+      67,
+      68,
+      69,
+      70,
+      71,
+      72,
+      73,
+      74,
+      75,
+      76,
+      77,
+      78,
+      79,
+      80,
+      81,
+      82,
+      83,
+      84,
+      85,
+      86,
+      87,
+      88,
+      89,
+      90,
+      91,
+      92,
+      93,
+      94,
+      95,
+      96,
+      97,
+      98,
+      99
+    ],
+    "observation_dimension": 1,
+    "ode": true,
+    "particles": 10000,
+    "swd_projections": 1000,
+    "terminal_time": 3.0
+  },
+  "execution": {
+    "elapsed_s": 3937.3646077919984,
+    "finished_at_utc": "2026-07-19T07:23:30.027035Z",
+    "jax_devices": [
+      "TFRT_CPU_0"
+    ],
+    "jax_version": "0.7.2",
+    "platform": {
+      "machine": "arm64",
+      "release": "25.5.0",
+      "system": "Darwin"
+    },
+    "python_executable_name": "python",
+    "python_version": "3.13.14 (main, Jun 23 2026, 15:20:43) [Clang 22.1.3 ]",
+    "started_at_utc": "2026-07-19T06:17:52.538734Z"
+  },
+  "label": "100/100 source-faithful selected-configuration reproduction-compatible run",
+  "per_id_results": [
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.06891570937888941,
+        "sliced_wasserstein_l1": 0.09377322994986281
+      },
+      "mc_id": 0,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.1040802101148604,
+        "sliced_wasserstein_l1": 0.1141936577500182
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.03264922970816248,
+        "sliced_wasserstein_l1": 0.09548877442318275
+      },
+      "mc_id": 1,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.022600303692709725,
+        "sliced_wasserstein_l1": 0.09276878330938158
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.012047793621225154,
+        "sliced_wasserstein_l1": 0.06502579965803025
+      },
+      "mc_id": 2,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.005596881905338691,
+        "sliced_wasserstein_l1": 0.05904409471586796
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.020648448470086508,
+        "sliced_wasserstein_l1": 0.08870963335465994
+      },
+      "mc_id": 3,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.05116297439687086,
+        "sliced_wasserstein_l1": 0.10006138160113144
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.014426982131915986,
+        "sliced_wasserstein_l1": 0.060177782135357065
+      },
+      "mc_id": 4,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.01194154773307079,
+        "sliced_wasserstein_l1": 0.057109334743782846
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.0488239412169504,
+        "sliced_wasserstein_l1": 0.0831088724212703
+      },
+      "mc_id": 5,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.06271602368229345,
+        "sliced_wasserstein_l1": 0.09303696373674894
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.06283535001772204,
+        "sliced_wasserstein_l1": 0.07373427590119798
+      },
+      "mc_id": 6,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.01669189826874349,
+        "sliced_wasserstein_l1": 0.05669838936194826
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.03621921111754006,
+        "sliced_wasserstein_l1": 0.08021821303146537
+      },
+      "mc_id": 7,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.020051216331967728,
+        "sliced_wasserstein_l1": 0.06563150098737912
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.04558563959133777,
+        "sliced_wasserstein_l1": 0.09810503545345538
+      },
+      "mc_id": 8,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.04710291340381191,
+        "sliced_wasserstein_l1": 0.09261694216873077
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.018591467076089545,
+        "sliced_wasserstein_l1": 0.06538085226729047
+      },
+      "mc_id": 9,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.008565463164904566,
+        "sliced_wasserstein_l1": 0.06298573487285027
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.03482571646739699,
+        "sliced_wasserstein_l1": 0.05679967272934064
+      },
+      "mc_id": 10,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.016673352284000888,
+        "sliced_wasserstein_l1": 0.06251043883629538
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.02397993086589409,
+        "sliced_wasserstein_l1": 0.0810018748517188
+      },
+      "mc_id": 11,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.04326380989962982,
+        "sliced_wasserstein_l1": 0.08484258349307595
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.022528846872763013,
+        "sliced_wasserstein_l1": 0.07539072810363967
+      },
+      "mc_id": 12,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.0240350974671139,
+        "sliced_wasserstein_l1": 0.08655175825606867
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.0172505155954237,
+        "sliced_wasserstein_l1": 0.07533019877383342
+      },
+      "mc_id": 13,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.013925609952305039,
+        "sliced_wasserstein_l1": 0.07191291804028546
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.13674925583503703,
+        "sliced_wasserstein_l1": 0.13425807430669245
+      },
+      "mc_id": 14,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.19217005082645844,
+        "sliced_wasserstein_l1": 0.15165201156090122
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.0122310681273865,
+        "sliced_wasserstein_l1": 0.06300401680423787
+      },
+      "mc_id": 15,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.00874067511443533,
+        "sliced_wasserstein_l1": 0.07584196783089482
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.09554554357930481,
+        "sliced_wasserstein_l1": 0.11836820981352218
+      },
+      "mc_id": 16,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.34224144904965786,
+        "sliced_wasserstein_l1": 0.19381190521839362
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.03355242019153578,
+        "sliced_wasserstein_l1": 0.07920532715009895
+      },
+      "mc_id": 17,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.03694658745061967,
+        "sliced_wasserstein_l1": 0.0815834985050807
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.04181830426004726,
+        "sliced_wasserstein_l1": 0.06601850317598792
+      },
+      "mc_id": 18,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.01563539731793071,
+        "sliced_wasserstein_l1": 0.055248678588276326
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.015516074232818578,
+        "sliced_wasserstein_l1": 0.0638787566633529
+      },
+      "mc_id": 19,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.016798210362882815,
+        "sliced_wasserstein_l1": 0.0709559686498657
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.03622937651844619,
+        "sliced_wasserstein_l1": 0.0839791851744826
+      },
+      "mc_id": 20,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.04667016577797182,
+        "sliced_wasserstein_l1": 0.08796245805616663
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.010971503994763232,
+        "sliced_wasserstein_l1": 0.07441394700956039
+      },
+      "mc_id": 21,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.013080369670669775,
+        "sliced_wasserstein_l1": 0.07989432486039345
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.016168654318198274,
+        "sliced_wasserstein_l1": 0.06457886736421524
+      },
+      "mc_id": 22,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.0049919676390353315,
+        "sliced_wasserstein_l1": 0.0848852300161524
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.036319512769618714,
+        "sliced_wasserstein_l1": 0.10049743455881818
+      },
+      "mc_id": 23,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.005085532129554061,
+        "sliced_wasserstein_l1": 0.0719266849790782
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.05515254441039166,
+        "sliced_wasserstein_l1": 0.12041122587202396
+      },
+      "mc_id": 24,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.040459661992828404,
+        "sliced_wasserstein_l1": 0.11297896203095471
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.048811202900908526,
+        "sliced_wasserstein_l1": 0.10304247915205773
+      },
+      "mc_id": 25,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.06572107317392184,
+        "sliced_wasserstein_l1": 0.11913945526796399
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.03188240598092994,
+        "sliced_wasserstein_l1": 0.08014028312665263
+      },
+      "mc_id": 26,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.039137608166440845,
+        "sliced_wasserstein_l1": 0.09522486774760414
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.10602543810555018,
+        "sliced_wasserstein_l1": 0.11829297198983159
+      },
+      "mc_id": 27,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.03694724866204143,
+        "sliced_wasserstein_l1": 0.07651384388479254
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.04519603577011913,
+        "sliced_wasserstein_l1": 0.07916126367801211
+      },
+      "mc_id": 28,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.021939693207344717,
+        "sliced_wasserstein_l1": 0.07377469620929651
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.021689055008637093,
+        "sliced_wasserstein_l1": 0.08072295588996176
+      },
+      "mc_id": 29,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.056758870853395885,
+        "sliced_wasserstein_l1": 0.07442207840830083
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.009695508118593954,
+        "sliced_wasserstein_l1": 0.06051299052441679
+      },
+      "mc_id": 30,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.006350280596347399,
+        "sliced_wasserstein_l1": 0.05894358104667147
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.009178384000445463,
+        "sliced_wasserstein_l1": 0.051127440029858966
+      },
+      "mc_id": 31,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.019952473956209704,
+        "sliced_wasserstein_l1": 0.06425940099215953
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.003946685433082616,
+        "sliced_wasserstein_l1": 0.055551979709069686
+      },
+      "mc_id": 32,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.00328472168500068,
+        "sliced_wasserstein_l1": 0.04778862503625366
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.027655494001120773,
+        "sliced_wasserstein_l1": 0.07749391586406214
+      },
+      "mc_id": 33,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.03421265167904036,
+        "sliced_wasserstein_l1": 0.08697936042638041
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.028753387199250597,
+        "sliced_wasserstein_l1": 0.07692327137757905
+      },
+      "mc_id": 34,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.01993589734175196,
+        "sliced_wasserstein_l1": 0.07640535980157984
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.13645953398436217,
+        "sliced_wasserstein_l1": 0.1364480284383062
+      },
+      "mc_id": 35,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.029749125474921145,
+        "sliced_wasserstein_l1": 0.10271344324946956
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.06378146133236005,
+        "sliced_wasserstein_l1": 0.09235928033253747
+      },
+      "mc_id": 36,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.04812120865442554,
+        "sliced_wasserstein_l1": 0.086689384262429
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.07318255498406302,
+        "sliced_wasserstein_l1": 0.10767117494804813
+      },
+      "mc_id": 37,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.017561540182111854,
+        "sliced_wasserstein_l1": 0.06834652722068182
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.008273956530541294,
+        "sliced_wasserstein_l1": 0.05691306423402274
+      },
+      "mc_id": 38,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.009875563619541753,
+        "sliced_wasserstein_l1": 0.05615763307782245
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.0165144041097703,
+        "sliced_wasserstein_l1": 0.0605974542095017
+      },
+      "mc_id": 39,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.0138462435193336,
+        "sliced_wasserstein_l1": 0.06122826874233936
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.016170796701220378,
+        "sliced_wasserstein_l1": 0.08013952435496892
+      },
+      "mc_id": 40,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.019334214268880462,
+        "sliced_wasserstein_l1": 0.09032684701612713
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.04916890253907513,
+        "sliced_wasserstein_l1": 0.08086656468931712
+      },
+      "mc_id": 41,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.0405525693366199,
+        "sliced_wasserstein_l1": 0.07788709644111658
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.007256297472821466,
+        "sliced_wasserstein_l1": 0.07348561540590808
+      },
+      "mc_id": 42,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.008433431257133975,
+        "sliced_wasserstein_l1": 0.07724571075033773
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.024409065411978917,
+        "sliced_wasserstein_l1": 0.06370226151295652
+      },
+      "mc_id": 43,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.019277530576053897,
+        "sliced_wasserstein_l1": 0.06291781723500348
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.040039119483248634,
+        "sliced_wasserstein_l1": 0.08278005955635163
+      },
+      "mc_id": 44,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.04469642354816345,
+        "sliced_wasserstein_l1": 0.09049253158975623
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.03512961263399872,
+        "sliced_wasserstein_l1": 0.059611055254663314
+      },
+      "mc_id": 45,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.046490006266494466,
+        "sliced_wasserstein_l1": 0.06558819056663613
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.11369175970179392,
+        "sliced_wasserstein_l1": 0.1278557554699578
+      },
+      "mc_id": 46,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.07983129343943343,
+        "sliced_wasserstein_l1": 0.11086546514285925
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.027697428658468883,
+        "sliced_wasserstein_l1": 0.07489820253231728
+      },
+      "mc_id": 47,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.04633156776296311,
+        "sliced_wasserstein_l1": 0.09180141884905282
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.020260159096109863,
+        "sliced_wasserstein_l1": 0.06779096363012993
+      },
+      "mc_id": 48,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.024386112175089368,
+        "sliced_wasserstein_l1": 0.07036352248847726
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.03944531135601915,
+        "sliced_wasserstein_l1": 0.07618845992022413
+      },
+      "mc_id": 49,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.0452475838479419,
+        "sliced_wasserstein_l1": 0.08156071506723642
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.0360583618084263,
+        "sliced_wasserstein_l1": 0.07701444707699175
+      },
+      "mc_id": 50,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.031100423984830886,
+        "sliced_wasserstein_l1": 0.08079768431036741
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.05991537505573065,
+        "sliced_wasserstein_l1": 0.08239907581648087
+      },
+      "mc_id": 51,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.04150528147417784,
+        "sliced_wasserstein_l1": 0.0803286513051208
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.05447490788475813,
+        "sliced_wasserstein_l1": 0.11760021948069026
+      },
+      "mc_id": 52,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.06253264799261019,
+        "sliced_wasserstein_l1": 0.1295152516090122
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.07492358212512154,
+        "sliced_wasserstein_l1": 0.12687321197892554
+      },
+      "mc_id": 53,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.08969898665002002,
+        "sliced_wasserstein_l1": 0.14107296116830592
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.0707384572987245,
+        "sliced_wasserstein_l1": 0.09114823366200502
+      },
+      "mc_id": 54,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.033805564988371554,
+        "sliced_wasserstein_l1": 0.07676588265000124
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.04127102011331055,
+        "sliced_wasserstein_l1": 0.06943578970690761
+      },
+      "mc_id": 55,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.07783626967561276,
+        "sliced_wasserstein_l1": 0.09177087269424802
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.04627221017489215,
+        "sliced_wasserstein_l1": 0.09637893817927559
+      },
+      "mc_id": 56,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.05869791245754579,
+        "sliced_wasserstein_l1": 0.10857449404500859
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.011626618670458364,
+        "sliced_wasserstein_l1": 0.06345020088313694
+      },
+      "mc_id": 57,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.011099062023747894,
+        "sliced_wasserstein_l1": 0.0691410606798498
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.007424533181744994,
+        "sliced_wasserstein_l1": 0.0803660057570237
+      },
+      "mc_id": 58,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.007724146789577309,
+        "sliced_wasserstein_l1": 0.07432703167673048
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.008910396477103507,
+        "sliced_wasserstein_l1": 0.08094452577129212
+      },
+      "mc_id": 59,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.016106152644097267,
+        "sliced_wasserstein_l1": 0.062215840517573784
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.00969159835821405,
+        "sliced_wasserstein_l1": 0.06230479728127258
+      },
+      "mc_id": 60,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.005869116657438355,
+        "sliced_wasserstein_l1": 0.058596985344374464
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.019176009155429566,
+        "sliced_wasserstein_l1": 0.06866662570424405
+      },
+      "mc_id": 61,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.005135600131749514,
+        "sliced_wasserstein_l1": 0.05307635624837906
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.032940480611043585,
+        "sliced_wasserstein_l1": 0.08115299207424688
+      },
+      "mc_id": 62,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.03482372699497808,
+        "sliced_wasserstein_l1": 0.09140849279706548
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.03197749116872859,
+        "sliced_wasserstein_l1": 0.06809506594495815
+      },
+      "mc_id": 63,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.03236664579919469,
+        "sliced_wasserstein_l1": 0.07776943302317336
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.058113540386713576,
+        "sliced_wasserstein_l1": 0.10809009416849835
+      },
+      "mc_id": 64,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.017437586287233672,
+        "sliced_wasserstein_l1": 0.08371655260944777
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.03705517679568902,
+        "sliced_wasserstein_l1": 0.0666989847928869
+      },
+      "mc_id": 65,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.04692990151171835,
+        "sliced_wasserstein_l1": 0.07384877337985353
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.008689254403217456,
+        "sliced_wasserstein_l1": 0.08486668168434311
+      },
+      "mc_id": 66,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.016076782830448687,
+        "sliced_wasserstein_l1": 0.09768473404989733
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.07126645197188654,
+        "sliced_wasserstein_l1": 0.11040639255924382
+      },
+      "mc_id": 67,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.06110364304780909,
+        "sliced_wasserstein_l1": 0.11514817619088384
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.03624296774386637,
+        "sliced_wasserstein_l1": 0.08161645072870102
+      },
+      "mc_id": 68,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.06364797838576038,
+        "sliced_wasserstein_l1": 0.10048007265920597
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.08115269530522308,
+        "sliced_wasserstein_l1": 0.1078712021514446
+      },
+      "mc_id": 69,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.10383150623922409,
+        "sliced_wasserstein_l1": 0.12228691157708348
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.012047078097153321,
+        "sliced_wasserstein_l1": 0.05282802747712897
+      },
+      "mc_id": 70,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.004723957569919281,
+        "sliced_wasserstein_l1": 0.05245945999016524
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.05991785078648852,
+        "sliced_wasserstein_l1": 0.11491187262007564
+      },
+      "mc_id": 71,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.056891032796029604,
+        "sliced_wasserstein_l1": 0.10992493726748974
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.01971028811507647,
+        "sliced_wasserstein_l1": 0.06904640752859757
+      },
+      "mc_id": 72,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.02313268783627494,
+        "sliced_wasserstein_l1": 0.07200333335528362
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.026023725980347207,
+        "sliced_wasserstein_l1": 0.07381699029228993
+      },
+      "mc_id": 73,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.04375853876538683,
+        "sliced_wasserstein_l1": 0.09402276074664541
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.03041544705388763,
+        "sliced_wasserstein_l1": 0.071578473069758
+      },
+      "mc_id": 74,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.010722311770555245,
+        "sliced_wasserstein_l1": 0.05720229914771131
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.021368221601264583,
+        "sliced_wasserstein_l1": 0.060236511044311235
+      },
+      "mc_id": 75,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.017955033908617683,
+        "sliced_wasserstein_l1": 0.06404413360656891
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.006374454683134088,
+        "sliced_wasserstein_l1": 0.05195730426191277
+      },
+      "mc_id": 76,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.004612117467660425,
+        "sliced_wasserstein_l1": 0.05257080446901189
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.04862932964889803,
+        "sliced_wasserstein_l1": 0.08538747543050669
+      },
+      "mc_id": 77,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.03398615226809264,
+        "sliced_wasserstein_l1": 0.07894425544411471
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.01833601858478738,
+        "sliced_wasserstein_l1": 0.05853182149447343
+      },
+      "mc_id": 78,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.011370484772635723,
+        "sliced_wasserstein_l1": 0.05796323886743226
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.014526385870776577,
+        "sliced_wasserstein_l1": 0.06762679215307542
+      },
+      "mc_id": 79,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.014778310033297956,
+        "sliced_wasserstein_l1": 0.061937744770012515
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.05530392966862325,
+        "sliced_wasserstein_l1": 0.09552879611228021
+      },
+      "mc_id": 80,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.014674989269232046,
+        "sliced_wasserstein_l1": 0.07155034906885474
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.01602442842946189,
+        "sliced_wasserstein_l1": 0.04950326153086329
+      },
+      "mc_id": 81,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.005394143010299702,
+        "sliced_wasserstein_l1": 0.04948276006101239
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.020363014436935393,
+        "sliced_wasserstein_l1": 0.0758942265053605
+      },
+      "mc_id": 82,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.024251092777101812,
+        "sliced_wasserstein_l1": 0.07456628404458072
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.018812894105280482,
+        "sliced_wasserstein_l1": 0.07049352945415321
+      },
+      "mc_id": 83,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.01686249745404021,
+        "sliced_wasserstein_l1": 0.07394344697611276
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.0977775590009634,
+        "sliced_wasserstein_l1": 0.1219960333594908
+      },
+      "mc_id": 84,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.1613756707509844,
+        "sliced_wasserstein_l1": 0.14869578800183555
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.010649266940816237,
+        "sliced_wasserstein_l1": 0.055396204483944736
+      },
+      "mc_id": 85,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.015503516562891954,
+        "sliced_wasserstein_l1": 0.06540279556958378
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.08835662877012126,
+        "sliced_wasserstein_l1": 0.12000226148366587
+      },
+      "mc_id": 86,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.09066643615092282,
+        "sliced_wasserstein_l1": 0.12378738113761353
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.006280290751351108,
+        "sliced_wasserstein_l1": 0.049275986886894586
+      },
+      "mc_id": 87,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.0075425721554638805,
+        "sliced_wasserstein_l1": 0.051205684022493016
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.08676744779219941,
+        "sliced_wasserstein_l1": 0.10086046158089164
+      },
+      "mc_id": 88,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.08672420842532068,
+        "sliced_wasserstein_l1": 0.10362834973472226
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.050064377262750294,
+        "sliced_wasserstein_l1": 0.06726747977149947
+      },
+      "mc_id": 89,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.025722523807794665,
+        "sliced_wasserstein_l1": 0.059359619647393745
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.022933841608613423,
+        "sliced_wasserstein_l1": 0.07734549062706542
+      },
+      "mc_id": 90,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.027591621165413194,
+        "sliced_wasserstein_l1": 0.0752111822642114
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.013372240585768063,
+        "sliced_wasserstein_l1": 0.05957422928911832
+      },
+      "mc_id": 91,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.014893498026516227,
+        "sliced_wasserstein_l1": 0.06704279968011366
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.013702302856228709,
+        "sliced_wasserstein_l1": 0.06260278890938169
+      },
+      "mc_id": 92,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.04625965636759917,
+        "sliced_wasserstein_l1": 0.09541376580991753
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.011406129241299162,
+        "sliced_wasserstein_l1": 0.06773922158955231
+      },
+      "mc_id": 93,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.009615428618909539,
+        "sliced_wasserstein_l1": 0.058359290651428225
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.025135606985714654,
+        "sliced_wasserstein_l1": 0.07027188348914146
+      },
+      "mc_id": 94,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.04735402475865002,
+        "sliced_wasserstein_l1": 0.08490848851766643
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.011799847063263463,
+        "sliced_wasserstein_l1": 0.06747323266203388
+      },
+      "mc_id": 95,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.024608777522026684,
+        "sliced_wasserstein_l1": 0.08356864837774786
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.020370931124455607,
+        "sliced_wasserstein_l1": 0.0796056202940244
+      },
+      "mc_id": 96,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.00661146014516815,
+        "sliced_wasserstein_l1": 0.06836157727901913
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.13167805524514758,
+        "sliced_wasserstein_l1": 0.13845705348834103
+      },
+      "mc_id": 97,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.1127275074632845,
+        "sliced_wasserstein_l1": 0.13545506362319115
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.005631224288311868,
+        "sliced_wasserstein_l1": 0.05855718767133343
+      },
+      "mc_id": 98,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.005273737343580874,
+        "sliced_wasserstein_l1": 0.054645480902919465
+      }
+    },
+    {
+      "diffusion_resampling": {
+        "posterior_mean_residual_squared_l2": 0.023480566420237135,
+        "sliced_wasserstein_l1": 0.07257987027150435
+      },
+      "mc_id": 99,
+      "multinomial_baseline": {
+        "posterior_mean_residual_squared_l2": 0.031187830388583963,
+        "sliced_wasserstein_l1": 0.07481256106368367
+      }
+    }
+  ],
+  "protocol_note": "The initial Euler T=2, K=8 partial pilot is retained separately. This run changes only terminal_time, diffusion_steps, and integrator to the paper-selected configuration.",
+  "reproduction_compatible": true,
+  "source_commit": "767effe3e755067eb8a04422597fbf37eb8ab754",
+  "source_paths": [
+    "experiments/run_gms.sh",
+    "experiments/gms/diffusion.py",
+    "experiments/gms/baselines.py",
+    "experiments/summary/print_gms_errs.py",
+    "experiments/rnd_keys.npy"
+  ],
+  "source_provenance": {
+    "author_source_sha256": {
+      "diffres/integators.py": "4cbcdba683cf27532b21acfa9c3295354c52cc2f4494b4063a0459629eb72954",
+      "diffres/resampling.py": "12b33fe23e1f62e18909cbcff9fa45db36ed65f6f943668728c9fe61795aa06e",
+      "diffres/tools.py": "54c60e1ac4f484532f51be57a63aacbdfb7a53ef8cef621495176ea7a843a5df",
+      "experiments/gms/baselines.py": "7b258df0ab4a7ca8eb5c91f6d5efbabbe359e0130db1cb9e4d3263a9241f1d5c",
+      "experiments/gms/diffusion.py": "886496bab31d254ac8030286ea350e48db817bd680cccc8cdda50c58a94ba774",
+      "experiments/rnd_keys.npy": "6e60685445936e0380220f34fc3fd15b5b7720070dc8c8f7f1148575de4ff5df",
+      "experiments/run_gms.sh": "c01a4c88a3ef05d6a5ffb585b9d4233b1b1a9e8c1c46d40360eab44963b3eee5",
+      "experiments/summary/print_gms_errs.py": "1df81cadaaf35ca5b26e8bd1ce6fc19aa4a60a2f7e161075087e518b3a0d17cd"
+    },
+    "expected_source_commit": "767effe3e755067eb8a04422597fbf37eb8ab754",
+    "expected_source_origin": "https://github.com/zgbkdlm/diffres.git",
+    "git_metadata_available": true,
+    "observed_remote_matches_expected": true,
+    "observed_source_commit": "767effe3e755067eb8a04422597fbf37eb8ab754",
+    "staged_diff_paths": [],
+    "tracked_worktree_diff_paths": []
+  },
+  "study_runtime_sha256": {
+    "study/run_gms_partial.py": "f8bd08d01ddc2fb8837aa5bd5f1788ea7cf721deae5d08db3234d18ede25091c",
+    "study/run_gms_selected_partial.py": "93da9132e53b5aefff85b9257a6b84dcf4f966bcce9e488f01ca577a7a747e0d"
+  }
+}

evidence/test_verify_stored_gms_evidence.py

@@ -0,0 +1,84 @@
+"""Focused standard-library tests for the self-contained stored-evidence verifier."""
+
+from __future__ import annotations
+
+import copy
+import json
+from pathlib import Path
+import tempfile
+import unittest
+
+import verify_stored_gms_evidence as verifier
+
+
+class StoredEvidenceVerifierTests(unittest.TestCase):
+    def test_canonical_record_passes(self) -> None:
+        report = verifier.verify_evidence(verifier.DEFAULT_INPUT)
+
+        self.assertEqual(report["record_count"], 100)
+        self.assertEqual(
+            report["diffusion_lower_is_better_wins"],
+            {
+                "sliced_wasserstein_l1": 54,
+                "posterior_mean_residual_squared_l2": 51,
+            },
+        )
+
+    def test_duplicate_json_key_is_rejected(self) -> None:
+        with tempfile.TemporaryDirectory() as directory:
+            path = Path(directory) / "duplicate.json"
+            path.write_text('{"configuration": {}, "configuration": {}}\n', encoding="utf-8")
+
+            with self.assertRaisesRegex(verifier.VerificationError, "duplicate JSON object key"):
+                verifier.strict_json_load(path)
+
+    def test_nonfinite_json_constant_is_rejected(self) -> None:
+        with tempfile.TemporaryDirectory() as directory:
+            path = Path(directory) / "nonfinite.json"
+            path.write_text('{"value": NaN}\n', encoding="utf-8")
+
+            with self.assertRaisesRegex(verifier.VerificationError, "non-standard JSON constant"):
+                verifier.strict_json_load(path)
+
+    def test_tampered_aggregate_is_rejected(self) -> None:
+        payload = json.loads(verifier.DEFAULT_INPUT.read_text(encoding="utf-8"))
+        tampered = copy.deepcopy(payload)
+        tampered["aggregate"]["diffusion_resampling"]["sliced_wasserstein_l1"]["mean"] += 0.01
+
+        with tempfile.TemporaryDirectory() as directory:
+            path = Path(directory) / "tampered.json"
+            path.write_text(json.dumps(tampered), encoding="utf-8")
+
+            with self.assertRaisesRegex(
+                verifier.VerificationError,
+                "aggregate.diffusion_resampling.sliced_wasserstein_l1.mean",
+            ):
+                verifier.verify_evidence(path)
+
+    def test_tampered_configuration_is_rejected(self) -> None:
+        payload = json.loads(verifier.DEFAULT_INPUT.read_text(encoding="utf-8"))
+        tampered = copy.deepcopy(payload)
+        tampered["configuration"]["particles"] = 9_999
+
+        with tempfile.TemporaryDirectory() as directory:
+            path = Path(directory) / "wrong-configuration.json"
+            path.write_text(json.dumps(tampered), encoding="utf-8")
+
+            with self.assertRaisesRegex(verifier.VerificationError, "configuration.particles"):
+                verifier.verify_evidence(path)
+
+    def test_out_of_order_id_is_rejected(self) -> None:
+        payload = json.loads(verifier.DEFAULT_INPUT.read_text(encoding="utf-8"))
+        tampered = copy.deepcopy(payload)
+        tampered["per_id_results"][0]["mc_id"] = 1
+
+        with tempfile.TemporaryDirectory() as directory:
+            path = Path(directory) / "wrong-id.json"
+            path.write_text(json.dumps(tampered), encoding="utf-8")
+
+            with self.assertRaisesRegex(verifier.VerificationError, "per_id_results\\[0\\].mc_id"):
+                verifier.verify_evidence(path)
+
+
+if __name__ == "__main__":
+    unittest.main()

evidence/validate_gms_selected_result.py

@@ -0,0 +1,430 @@
+"""Independently validate the recorded 100-ID selected GMS result.
+
+This is a post-hoc integrity and aggregation check.  It deliberately does not
+import the numerical runner or re-execute JAX: its purpose is to make the
+recorded result's input identity, raw values, and reported aggregates
+independently traceable from a small standard-Python programme.
+"""
+
+from __future__ import annotations
+
+import argparse
+import datetime as dt
+import hashlib
+import json
+import math
+import os
+from pathlib import Path
+import platform
+import statistics
+import subprocess
+import sys
+import tempfile
+from typing import Any, Iterable, NoReturn
+
+import scipy
+from scipy.stats import t as student_t
+
+
+REPOSITORY_ROOT = Path(__file__).resolve().parents[1]
+STUDY_ROOT = REPOSITORY_ROOT / "study"
+DEFAULT_INPUT = STUDY_ROOT / "gms-selected-ids-0-99.json"
+DEFAULT_OUTPUT = STUDY_ROOT / "gms-selected-ids-0-99-validation.json"
+SOURCE_COMMIT = "767effe3e755067eb8a04422597fbf37eb8ab754"
+EXPECTED_SOURCE_ORIGIN = "https://github.com/zgbkdlm/diffres.git"
+EXPECTED_CONFIGURATION = {
+    "diffusion_a": -1.0,
+    "terminal_time": 3.0,
+    "diffusion_steps": 128,
+    "integrator": "jentzen_and_kloeden",
+    "ode": True,
+    "particles": 10_000,
+    "components": 5,
+    "dimension": 8,
+    "observation_dimension": 1,
+    "swd_projections": 1_000,
+}
+METHODS = ("diffusion_resampling", "multinomial_baseline")
+METRICS = ("sliced_wasserstein_l1", "posterior_mean_residual_squared_l2")
+
+
+class ValidationError(ValueError):
+    """Raised when the recorded result is not internally traceable."""
+
+
+def fail(message: str) -> NoReturn:
+    raise ValidationError(message)
+
+
+def sha256_file(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as handle:
+        for block in iter(lambda: handle.read(1024 * 1024), b""):
+            digest.update(block)
+    return digest.hexdigest()
+
+
+def utc_now() -> str:
+    return dt.datetime.now(tz=dt.timezone.utc).isoformat().replace("+00:00", "Z")
+
+
+def strict_json_load(path: Path) -> dict[str, Any]:
+    def reject_nonfinite(token: str) -> NoReturn:
+        fail(f"{path.name} contains non-standard JSON constant {token!r}")
+
+    def reject_duplicate_keys(pairs: list[tuple[str, Any]]) -> dict[str, Any]:
+        object_value: dict[str, Any] = {}
+        for key, value in pairs:
+            if key in object_value:
+                fail(f"{path.name} contains duplicate JSON object key {key!r}")
+            object_value[key] = value
+        return object_value
+
+    try:
+        value = json.loads(
+            path.read_text(encoding="utf-8"),
+            parse_constant=reject_nonfinite,
+            object_pairs_hook=reject_duplicate_keys,
+        )
+    except (OSError, json.JSONDecodeError) as error:
+        fail(f"Cannot parse strict JSON from {path}: {error}")
+    if not isinstance(value, dict):
+        fail(f"Top-level value in {path.name} must be an object")
+    return value
+
+
+def git_text(*arguments: str) -> str:
+    completed = subprocess.run(
+        ["git", *arguments],
+        cwd=REPOSITORY_ROOT,
+        check=False,
+        capture_output=True,
+        text=True,
+    )
+    if completed.returncode != 0:
+        fail(f"Git command failed: git {' '.join(arguments)}: {completed.stderr.strip()}")
+    return completed.stdout.strip()
+
+
+def resolve_within_repository(relative_path: str) -> Path:
+    candidate = (REPOSITORY_ROOT / relative_path).resolve()
+    try:
+        candidate.relative_to(REPOSITORY_ROOT.resolve())
+    except ValueError:
+        fail(f"Recorded source path escapes repository root: {relative_path}")
+    return candidate
+
+
+def require_finite_number(value: object, path: str) -> float:
+    if isinstance(value, bool) or not isinstance(value, (int, float)):
+        fail(f"{path} must be a numeric JSON value")
+    converted = float(value)
+    if not math.isfinite(converted):
+        fail(f"{path} must be finite")
+    return converted
+
+
+def require_equal(actual: object, expected: object, path: str) -> None:
+    if actual != expected:
+        fail(f"{path} is {actual!r}; expected {expected!r}")
+
+
+def require_close(actual: object, expected: float, path: str) -> None:
+    value = require_finite_number(actual, path)
+    if not math.isclose(value, expected, rel_tol=0.0, abs_tol=1e-12):
+        fail(f"{path} is {value:.17g}; recomputation is {expected:.17g}")
+
+
+def summary(values: Iterable[float]) -> dict[str, float]:
+    values_list = list(values)
+    if not values_list:
+        fail("Cannot aggregate an empty metric series")
+    return {
+        "mean": statistics.fmean(values_list),
+        "population_standard_deviation": statistics.pstdev(values_list),
+        "minimum": min(values_list),
+        "maximum": max(values_list),
+    }
+
+
+def paired_t_interval(values: Iterable[float]) -> dict[str, float | int]:
+    values_list = list(values)
+    if len(values_list) < 2:
+        fail("A paired t interval needs at least two observations")
+    mean = statistics.fmean(values_list)
+    sample_standard_deviation = statistics.stdev(values_list)
+    standard_error = sample_standard_deviation / math.sqrt(len(values_list))
+    critical_value = float(student_t.ppf(0.975, len(values_list) - 1))
+    half_width = critical_value * standard_error
+    return {
+        "sample_count": len(values_list),
+        "mean": mean,
+        "sample_standard_deviation": sample_standard_deviation,
+        "standard_error": standard_error,
+        "critical_value": critical_value,
+        "lower": mean - half_width,
+        "upper": mean + half_width,
+    }
+
+
+def validate_source_identity(result: dict[str, Any]) -> dict[str, Any]:
+    provenance = result.get("source_provenance")
+    if not isinstance(provenance, dict):
+        fail("source_provenance must be an object")
+
+    require_equal(result.get("source_commit"), SOURCE_COMMIT, "source_commit")
+    require_equal(
+        provenance.get("expected_source_commit"), SOURCE_COMMIT, "source_provenance.expected_source_commit"
+    )
+    require_equal(
+        provenance.get("observed_source_commit"), SOURCE_COMMIT, "source_provenance.observed_source_commit"
+    )
+    require_equal(
+        provenance.get("expected_source_origin"),
+        EXPECTED_SOURCE_ORIGIN,
+        "source_provenance.expected_source_origin",
+    )
+    require_equal(
+        provenance.get("observed_remote_matches_expected"),
+        True,
+        "source_provenance.observed_remote_matches_expected",
+    )
+    require_equal(provenance.get("git_metadata_available"), True, "source_provenance.git_metadata_available")
+    require_equal(
+        provenance.get("tracked_worktree_diff_paths"), [], "source_provenance.tracked_worktree_diff_paths"
+    )
+    require_equal(provenance.get("staged_diff_paths"), [], "source_provenance.staged_diff_paths")
+
+    current_commit = git_text("rev-parse", "HEAD")
+    current_origin = git_text("config", "--get", "remote.origin.url")
+    current_diff = git_text("diff", "--name-only")
+    current_staged_diff = git_text("diff", "--cached", "--name-only")
+    require_equal(current_commit, SOURCE_COMMIT, "current Git commit")
+    require_equal(current_origin, EXPECTED_SOURCE_ORIGIN, "current Git origin")
+    require_equal(current_diff, "", "current tracked worktree diff")
+    require_equal(current_staged_diff, "", "current staged worktree diff")
+
+    recorded_hashes = provenance.get("author_source_sha256")
+    if not isinstance(recorded_hashes, dict) or not recorded_hashes:
+        fail("source_provenance.author_source_sha256 must be a non-empty object")
+    checked_hashes: dict[str, str] = {}
+    for relative_path, recorded_hash in sorted(recorded_hashes.items()):
+        if not isinstance(relative_path, str) or not isinstance(recorded_hash, str):
+            fail("source_provenance.author_source_sha256 must map strings to strings")
+        path = resolve_within_repository(relative_path)
+        if not path.is_file():
+            fail(f"Recorded source file is unavailable: {relative_path}")
+        actual_hash = sha256_file(path)
+        require_equal(actual_hash, recorded_hash, f"SHA-256 for {relative_path}")
+        checked_hashes[relative_path] = actual_hash
+
+    return {
+        "current_commit": current_commit,
+        "current_origin_matches_expected": current_origin == EXPECTED_SOURCE_ORIGIN,
+        "current_tracked_worktree_clean": current_diff == "" and current_staged_diff == "",
+        "checked_author_source_sha256": checked_hashes,
+    }
+
+
+def validate_protocol(result: dict[str, Any]) -> None:
+    require_equal(result.get("reproduction_compatible"), True, "reproduction_compatible")
+    configuration = result.get("configuration")
+    if not isinstance(configuration, dict):
+        fail("configuration must be an object")
+    for key, expected in EXPECTED_CONFIGURATION.items():
+        require_equal(configuration.get(key), expected, f"configuration.{key}")
+    require_equal(configuration.get("mc_ids"), list(range(100)), "configuration.mc_ids")
+    require_equal(configuration.get("author_keys_path"), "experiments/rnd_keys.npy", "configuration.author_keys_path")
+
+    source_paths = result.get("source_paths")
+    required_paths = {
+        "experiments/run_gms.sh",
+        "experiments/gms/diffusion.py",
+        "experiments/gms/baselines.py",
+        "experiments/summary/print_gms_errs.py",
+        "experiments/rnd_keys.npy",
+    }
+    if not isinstance(source_paths, list) or not required_paths.issubset(source_paths):
+        fail("source_paths does not record the required released GMS files")
+
+
+def validate_records_and_aggregates(result: dict[str, Any]) -> dict[str, Any]:
+    records = result.get("per_id_results")
+    if not isinstance(records, list):
+        fail("per_id_results must be a list")
+    if len(records) != 100:
+        fail(f"per_id_results has {len(records)} records; expected 100")
+
+    series: dict[str, dict[str, list[float]]] = {
+        method: {metric: [] for metric in METRICS} for method in METHODS
+    }
+    for expected_id, record in enumerate(records):
+        if not isinstance(record, dict):
+            fail(f"per_id_results[{expected_id}] must be an object")
+        require_equal(record.get("mc_id"), expected_id, f"per_id_results[{expected_id}].mc_id")
+        for method in METHODS:
+            method_values = record.get(method)
+            if not isinstance(method_values, dict):
+                fail(f"per_id_results[{expected_id}].{method} must be an object")
+            for metric in METRICS:
+                series[method][metric].append(
+                    require_finite_number(
+                        method_values.get(metric), f"per_id_results[{expected_id}].{method}.{metric}"
+                    )
+                )
+
+    aggregate = result.get("aggregate")
+    if not isinstance(aggregate, dict):
+        fail("aggregate must be an object")
+    recomputed: dict[str, Any] = {}
+    for method in METHODS:
+        reported_method = aggregate.get(method)
+        if not isinstance(reported_method, dict):
+            fail(f"aggregate.{method} must be an object")
+        recomputed[method] = {}
+        for metric in METRICS:
+            reported_metric = reported_method.get(metric)
+            if not isinstance(reported_metric, dict):
+                fail(f"aggregate.{method}.{metric} must be an object")
+            computed = summary(series[method][metric])
+            for name, value in computed.items():
+                require_close(reported_metric.get(name), value, f"aggregate.{method}.{metric}.{name}")
+            recomputed[method][metric] = computed
+
+    reported_paired = aggregate.get("paired_diffusion_minus_multinomial")
+    if not isinstance(reported_paired, dict):
+        fail("aggregate.paired_diffusion_minus_multinomial must be an object")
+    recomputed_paired: dict[str, dict[str, float]] = {}
+    intervals: dict[str, dict[str, float | int]] = {}
+    wins: dict[str, int] = {}
+    for metric in METRICS:
+        values = [
+            diffusion - multinomial
+            for diffusion, multinomial in zip(
+                series["diffusion_resampling"][metric],
+                series["multinomial_baseline"][metric],
+                strict=True,
+            )
+        ]
+        reported_metric = reported_paired.get(metric)
+        if not isinstance(reported_metric, dict):
+            fail(f"aggregate.paired_diffusion_minus_multinomial.{metric} must be an object")
+        computed = summary(values)
+        for name, value in computed.items():
+            require_close(
+                reported_metric.get(name), value, f"aggregate.paired_diffusion_minus_multinomial.{metric}.{name}"
+            )
+        recomputed_paired[metric] = computed
+        intervals[metric] = paired_t_interval(values)
+        wins[metric] = sum(value < 0.0 for value in values)
+
+    reported_wins = aggregate.get("diffusion_lower_is_better_wins")
+    if not isinstance(reported_wins, dict):
+        fail("aggregate.diffusion_lower_is_better_wins must be an object")
+    for metric, count in wins.items():
+        require_equal(reported_wins.get(metric), count, f"aggregate.diffusion_lower_is_better_wins.{metric}")
+
+    return {
+        "record_count": len(records),
+        "recomputed_aggregate": recomputed,
+        "recomputed_paired_aggregate": recomputed_paired,
+        "paired_diffusion_minus_multinomial_95_percent_t_intervals": intervals,
+        "diffusion_lower_is_better_wins": wins,
+    }
+
+
+def validate_result(input_path: Path) -> dict[str, Any]:
+    result = strict_json_load(input_path)
+    validate_protocol(result)
+    source_identity = validate_source_identity(result)
+    aggregation = validate_records_and_aggregates(result)
+    return {
+        "schema_version": 1,
+        "validation_status": "passed",
+        "validation_type": "post-hoc integrity and aggregate recomputation; no numerical rerun",
+        "checked_at_utc": utc_now(),
+        "input": {
+            "path": input_path.relative_to(REPOSITORY_ROOT).as_posix(),
+            "sha256": sha256_file(input_path),
+        },
+        "checks": {
+            "strict_json": "passed",
+            "selected_protocol": "passed",
+            "current_source_identity_and_hashes": source_identity,
+            "raw_record_integrity_and_aggregate_recomputation": aggregation,
+        },
+        "validator": {
+            "path": Path(__file__).relative_to(REPOSITORY_ROOT).as_posix(),
+            "sha256": sha256_file(Path(__file__)),
+            "python_version": sys.version,
+            "scipy_version": scipy.__version__,
+            "platform": {
+                "system": platform.system(),
+                "release": platform.release(),
+                "machine": platform.machine(),
+            },
+        },
+        "claim_boundary": (
+            "This check independently recomputes the recorded aggregates and paired intervals "
+            "from the 100 raw JSON records while checking the current pinned source identity. "
+            "It does not rerun the numerical experiment, independently reproduce the paper, "
+            "or establish an outperformance or significance claim."
+        ),
+    }
+
+
+def resolve_study_path(requested: Path) -> Path:
+    path = requested if requested.is_absolute() else REPOSITORY_ROOT / requested
+    path = path.resolve()
+    try:
+        path.relative_to(STUDY_ROOT.resolve())
+    except ValueError:
+        fail(f"Path must be contained in {STUDY_ROOT}: {path}")
+    return path
+
+
+def write_json(path: Path, payload: dict[str, Any], *, overwrite: bool) -> None:
+    serialised = json.dumps(payload, indent=2, sort_keys=True, allow_nan=False) + "\n"
+    with tempfile.NamedTemporaryFile(
+        mode="w", encoding="utf-8", dir=path.parent, prefix=f".{path.name}.", delete=False
+    ) as temporary_file:
+        temporary_file.write(serialised)
+        temporary_file.flush()
+        os.fsync(temporary_file.fileno())
+        temporary_path = Path(temporary_file.name)
+    try:
+        if overwrite:
+            os.replace(temporary_path, path)
+        else:
+            os.link(temporary_path, path)
+    except FileExistsError as error:
+        raise FileExistsError(
+            f"Refusing to overwrite {path}; pass --overwrite for a deliberate rerun."
+        ) from error
+    finally:
+        temporary_path.unlink(missing_ok=True)
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("--input", type=Path, default=DEFAULT_INPUT)
+    parser.add_argument("--output", type=Path, default=DEFAULT_OUTPUT)
+    parser.add_argument("--overwrite", action="store_true")
+    return parser.parse_args()
+
+
+def main() -> None:
+    args = parse_args()
+    input_path = resolve_study_path(args.input)
+    output_path = resolve_study_path(args.output)
+    if output_path.exists() and not args.overwrite:
+        raise FileExistsError(
+            f"Refusing to overwrite {output_path}; pass --overwrite for a deliberate rerun."
+        )
+    payload = validate_result(input_path)
+    write_json(output_path, payload, overwrite=args.overwrite)
+    print(f"Wrote validation result to {output_path}")
+
+
+if __name__ == "__main__":
+    main()

evidence/verify_stored_gms_evidence.py

@@ -0,0 +1,343 @@
+"""Verify the structure and arithmetic of the stored 100-ID GMS evidence.
+
+This verifier uses only Python's standard library and never starts the numerical
+experiment, contacts a network service, opens a source checkout, or writes an
+output file. It checks the immutable JSON record's declared configuration,
+ordered records, and reported descriptive aggregates.
+
+It cannot establish that the recorded source hashes match a checkout, that the
+stored values came from a particular execution, or that the paper-level claims
+hold. Those are intentionally outside this file-only check.
+"""
+
+from __future__ import annotations
+
+import argparse
+import hashlib
+import json
+import math
+from pathlib import Path
+import statistics
+import sys
+from typing import Any, Iterable, NoReturn
+
+
+EVIDENCE_DIRECTORY = Path(__file__).resolve().parent
+DEFAULT_INPUT = EVIDENCE_DIRECTORY / "gms-selected-ids-0-99.json"
+METHODS = ("diffusion_resampling", "multinomial_baseline")
+METRICS = ("sliced_wasserstein_l1", "posterior_mean_residual_squared_l2")
+SUMMARY_FIELDS = ("mean", "population_standard_deviation", "minimum", "maximum")
+EXPECTED_CONFIGURATION = {
+    "author_keys_path": "experiments/rnd_keys.npy",
+    "components": 5,
+    "diffusion_a": -1.0,
+    "diffusion_steps": 128,
+    "dimension": 8,
+    "execution": "one CPU process; compiled functions shared across ids",
+    "integrator": "jentzen_and_kloeden",
+    "mc_ids": list(range(100)),
+    "observation_dimension": 1,
+    "ode": True,
+    "particles": 10_000,
+    "swd_projections": 1_000,
+    "terminal_time": 3.0,
+}
+EXPECTED_SOURCE_COMMIT = "767effe3e755067eb8a04422597fbf37eb8ab754"
+EXPECTED_TOP_LEVEL_KEYS = {
+    "aggregate",
+    "claim_boundary",
+    "configuration",
+    "execution",
+    "label",
+    "per_id_results",
+    "protocol_note",
+    "reproduction_compatible",
+    "source_commit",
+    "source_paths",
+    "source_provenance",
+    "study_runtime_sha256",
+}
+ABSOLUTE_TOLERANCE = 1e-12
+
+
+class VerificationError(ValueError):
+    """Raised when a stored evidence record is malformed or internally inconsistent."""
+
+
+def fail(message: str) -> NoReturn:
+    raise VerificationError(message)
+
+
+def sha256_file(path: Path) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as handle:
+        for block in iter(lambda: handle.read(1024 * 1024), b""):
+            digest.update(block)
+    return digest.hexdigest()
+
+
+def strict_json_load(path: Path) -> dict[str, Any]:
+    """Load a JSON object while rejecting duplicate keys and non-finite constants."""
+
+    def reject_nonfinite(token: str) -> NoReturn:
+        fail(f"{path.name} contains non-standard JSON constant {token!r}")
+
+    def reject_duplicate_keys(pairs: list[tuple[str, Any]]) -> dict[str, Any]:
+        value: dict[str, Any] = {}
+        for key, item in pairs:
+            if key in value:
+                fail(f"{path.name} contains duplicate JSON object key {key!r}")
+            value[key] = item
+        return value
+
+    try:
+        loaded = json.loads(
+            path.read_text(encoding="utf-8"),
+            parse_constant=reject_nonfinite,
+            object_pairs_hook=reject_duplicate_keys,
+        )
+    except (OSError, json.JSONDecodeError) as error:
+        fail(f"Cannot parse strict JSON from {path}: {error}")
+    if not isinstance(loaded, dict):
+        fail(f"Top-level value in {path.name} must be an object")
+    return loaded
+
+
+def require_object(value: object, path: str, *, keys: set[str] | None = None) -> dict[str, Any]:
+    if not isinstance(value, dict):
+        fail(f"{path} must be an object")
+    if keys is not None and set(value) != keys:
+        fail(f"{path} has keys {sorted(value)!r}; expected {sorted(keys)!r}")
+    return value
+
+
+def require_list(value: object, path: str) -> list[Any]:
+    if not isinstance(value, list):
+        fail(f"{path} must be a list")
+    return value
+
+
+def require_integer(value: object, path: str) -> int:
+    if isinstance(value, bool) or not isinstance(value, int):
+        fail(f"{path} must be an integer")
+    return value
+
+
+def require_number(value: object, path: str) -> float:
+    if isinstance(value, bool) or not isinstance(value, (int, float)):
+        fail(f"{path} must be a numeric JSON value")
+    number = float(value)
+    if not math.isfinite(number):
+        fail(f"{path} must be finite")
+    return number
+
+
+def require_equal(actual: object, expected: object, path: str) -> None:
+    if actual != expected:
+        fail(f"{path} is {actual!r}; expected {expected!r}")
+
+
+def require_close(actual: object, expected: float, path: str) -> None:
+    value = require_number(actual, path)
+    if not math.isclose(value, expected, rel_tol=0.0, abs_tol=ABSOLUTE_TOLERANCE):
+        fail(f"{path} is {value:.17g}; recomputation is {expected:.17g}")
+
+
+def summary(values: Iterable[float]) -> dict[str, float]:
+    series = list(values)
+    if not series:
+        fail("Cannot aggregate an empty metric series")
+    return {
+        "mean": statistics.fmean(series),
+        "population_standard_deviation": statistics.pstdev(series),
+        "minimum": min(series),
+        "maximum": max(series),
+    }
+
+
+def verify_declared_configuration(result: dict[str, Any]) -> None:
+    configuration = require_object(
+        result.get("configuration"), "configuration", keys=set(EXPECTED_CONFIGURATION)
+    )
+    for name, expected in EXPECTED_CONFIGURATION.items():
+        value = configuration[name]
+        path = f"configuration.{name}"
+        if name == "mc_ids":
+            identifiers = require_list(value, path)
+            require_equal(
+                [require_integer(item, f"{path}[{index}]") for index, item in enumerate(identifiers)],
+                expected,
+                path,
+            )
+            continue
+        if isinstance(expected, bool):
+            if not isinstance(value, bool):
+                fail(f"{path} must be a Boolean")
+        elif isinstance(expected, int):
+            require_integer(value, path)
+        elif isinstance(expected, float):
+            require_number(value, path)
+        require_equal(value, expected, path)
+
+    require_equal(result.get("source_commit"), EXPECTED_SOURCE_COMMIT, "source_commit declaration")
+
+
+def verify_records_and_aggregates(result: dict[str, Any]) -> dict[str, Any]:
+    records = require_list(result.get("per_id_results"), "per_id_results")
+    if len(records) != 100:
+        fail(f"per_id_results has {len(records)} records; expected 100")
+
+    series: dict[str, dict[str, list[float]]] = {
+        method: {metric: [] for metric in METRICS} for method in METHODS
+    }
+    for expected_id, record_value in enumerate(records):
+        record = require_object(
+            record_value,
+            f"per_id_results[{expected_id}]",
+            keys={"mc_id", *METHODS},
+        )
+        require_equal(
+            require_integer(record["mc_id"], f"per_id_results[{expected_id}].mc_id"),
+            expected_id,
+            f"per_id_results[{expected_id}].mc_id",
+        )
+        for method in METHODS:
+            values = require_object(
+                record[method],
+                f"per_id_results[{expected_id}].{method}",
+                keys=set(METRICS),
+            )
+            for metric in METRICS:
+                series[method][metric].append(
+                    require_number(
+                        values[metric], f"per_id_results[{expected_id}].{method}.{metric}"
+                    )
+                )
+
+    aggregate = require_object(
+        result.get("aggregate"),
+        "aggregate",
+        keys={*METHODS, "paired_diffusion_minus_multinomial", "diffusion_lower_is_better_wins"},
+    )
+    recomputed: dict[str, Any] = {}
+    for method in METHODS:
+        reported_method = require_object(
+            aggregate[method], f"aggregate.{method}", keys=set(METRICS)
+        )
+        recomputed[method] = {}
+        for metric in METRICS:
+            reported_summary = require_object(
+                reported_method[metric],
+                f"aggregate.{method}.{metric}",
+                keys=set(SUMMARY_FIELDS),
+            )
+            calculated = summary(series[method][metric])
+            for field, value in calculated.items():
+                require_close(
+                    reported_summary[field], value, f"aggregate.{method}.{metric}.{field}"
+                )
+            recomputed[method][metric] = calculated
+
+    paired_reported = require_object(
+        aggregate["paired_diffusion_minus_multinomial"],
+        "aggregate.paired_diffusion_minus_multinomial",
+        keys=set(METRICS),
+    )
+    wins_reported = require_object(
+        aggregate["diffusion_lower_is_better_wins"],
+        "aggregate.diffusion_lower_is_better_wins",
+        keys=set(METRICS),
+    )
+    paired: dict[str, Any] = {}
+    lower_counts: dict[str, int] = {}
+    for metric in METRICS:
+        differences = [
+            diffusion - multinomial
+            for diffusion, multinomial in zip(
+                series["diffusion_resampling"][metric],
+                series["multinomial_baseline"][metric],
+                strict=True,
+            )
+        ]
+        reported_summary = require_object(
+            paired_reported[metric],
+            f"aggregate.paired_diffusion_minus_multinomial.{metric}",
+            keys=set(SUMMARY_FIELDS),
+        )
+        calculated = summary(differences)
+        for field, value in calculated.items():
+            require_close(
+                reported_summary[field], value,
+                f"aggregate.paired_diffusion_minus_multinomial.{metric}.{field}",
+            )
+        lower_count = sum(value < 0.0 for value in differences)
+        require_equal(
+            require_integer(
+                wins_reported[metric], f"aggregate.diffusion_lower_is_better_wins.{metric}"
+            ),
+            lower_count,
+            f"aggregate.diffusion_lower_is_better_wins.{metric}",
+        )
+        paired[metric] = calculated
+        lower_counts[metric] = lower_count
+
+    return {
+        "record_count": len(records),
+        "recomputed_aggregate": recomputed,
+        "recomputed_paired_aggregate": paired,
+        "diffusion_lower_is_better_wins": lower_counts,
+    }
+
+
+def verify_evidence(input_path: Path) -> dict[str, Any]:
+    """Verify one stored result record and return the recomputed evidence summary."""
+
+    result = strict_json_load(input_path)
+    require_object(result, "top-level record", keys=EXPECTED_TOP_LEVEL_KEYS)
+    verify_declared_configuration(result)
+    recomputed = verify_records_and_aggregates(result)
+    return {
+        "input_sha256": sha256_file(input_path),
+        "declared_source_commit": EXPECTED_SOURCE_COMMIT,
+        **recomputed,
+    }
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument(
+        "--input",
+        type=Path,
+        default=DEFAULT_INPUT,
+        help="stored GMS JSON to inspect (default: evidence/gms-selected-ids-0-99.json)",
+    )
+    return parser.parse_args()
+
+
+def main() -> int:
+    args = parse_args()
+    try:
+        report = verify_evidence(args.input)
+    except (OSError, VerificationError) as error:
+        print(f"Stored-evidence verification: failed: {error}", file=sys.stderr)
+        return 1
+
+    print("Stored-evidence verification: passed")
+    print(f"Input SHA-256: {report['input_sha256']}")
+    print(f"Ordered records checked: {report['record_count']} (IDs 0–99)")
+    print("Methods checked: diffusion_resampling, multinomial_baseline")
+    print("Metrics checked: sliced_wasserstein_l1, posterior_mean_residual_squared_l2")
+    print(f"Diffusion-lower counts: {report['diffusion_lower_is_better_wins']}")
+    print(
+        "Recomputed from this file only: declared configuration, strict JSON structure, "
+        "raw-record aggregates, paired aggregates, and lower-counts."
+    )
+    print(
+        "Not established: source-checkout identity or hashes, numerical-execution provenance, "
+        "paper reproduction, statistical significance, or any outperformance claim."
+    )
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

index.html

@@ -1,19 +1,170 @@
 <!doctype html>
-<html>
-	<head>
-		<meta charset="utf-8" />
-		<meta name="viewport" content="width=device-width" />
-		<title>My static Space</title>
-		<link rel="stylesheet" href="style.css" />
-	</head>
-	<body>
-		<div class="card">
-			<h1>Welcome to your static Space!</h1>
-			<p>You can modify this app directly by editing <i>index.html</i> in the Files and versions tab.</p>
-			<p>
-				Also don't forget to check the
-				<a href="https://huggingface.co/docs/hub/spaces" target="_blank">Spaces documentation</a>.
-			</p>
-		</div>
-	</body>
+<html lang="en">
+  <head>
+    <meta charset="utf-8" />
+    <meta name="viewport" content="width=device-width, initial-scale=1" />
+    <meta
+      name="description"
+      content="Traceable historic 100-ID Gaussian-mixture evidence for Diffusion differentiable resampling."
+    />
+    <link rel="icon" href="data:," />
+    <title>DiffRes: traceable 100-ID Gaussian-mixture study</title>
+    <style>
+      :root {
+        color-scheme: light dark;
+        --paper: #f8fafc;
+        --ink: #172033;
+        --muted: #526076;
+        --panel: #ffffff;
+        --line: #d8e0ec;
+        --accent: #215bd8;
+        --warning: #7a4d00;
+        --warning-bg: #fff6e3;
+      }
+
+      @media (prefers-color-scheme: dark) {
+        :root {
+          --paper: #10151f;
+          --ink: #edf2ff;
+          --muted: #afbbce;
+          --panel: #182131;
+          --line: #2b384d;
+          --accent: #8bb2ff;
+          --warning: #ffce72;
+          --warning-bg: #2b2417;
+        }
+      }
+
+      * { box-sizing: border-box; }
+      body {
+        margin: 0;
+        background: var(--paper);
+        color: var(--ink);
+        font: 16px/1.55 system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif;
+      }
+      main { max-width: 900px; margin: 0 auto; padding: 56px 24px 72px; }
+      h1, h2 { line-height: 1.2; }
+      h1 { font-size: clamp(2rem, 5vw, 3.25rem); margin: 0.45rem 0 1rem; }
+      h2 { margin: 2.4rem 0 0.7rem; font-size: 1.3rem; }
+      p, li { max-width: 76ch; }
+      a { color: var(--accent); }
+      .eyebrow { color: var(--muted); font-weight: 700; letter-spacing: 0.08em; text-transform: uppercase; }
+      .status {
+        display: inline-block;
+        border: 1px solid var(--line);
+        border-radius: 999px;
+        padding: 0.2rem 0.65rem;
+        color: var(--muted);
+        font-size: 0.85rem;
+        font-weight: 650;
+      }
+      .card, .warning {
+        border: 1px solid var(--line);
+        border-radius: 12px;
+        padding: 1.1rem 1.2rem;
+        background: var(--panel);
+      }
+      .warning { background: var(--warning-bg); color: var(--warning); }
+      table { border-collapse: collapse; width: 100%; margin: 1rem 0; }
+      th, td { border-bottom: 1px solid var(--line); padding: 0.7rem 0.55rem; text-align: left; vertical-align: top; }
+      th:first-child, td:first-child { padding-left: 0; }
+      th:last-child, td:last-child { padding-right: 0; }
+      .number { font-variant-numeric: tabular-nums; text-align: right; white-space: nowrap; }
+      footer { border-top: 1px solid var(--line); color: var(--muted); font-size: 0.9rem; margin-top: 3rem; padding-top: 1rem; }
+    </style>
+  </head>
+  <body>
+    <main>
+      <div class="eyebrow">ICML 2026 evidence package</div>
+      <h1>DiffRes: traceable 100-ID Gaussian-mixture study</h1>
+      <p class="status">Historic stored output · static package · no numerical execution</p>
+      <p>
+        A local study wrapper using released implementation, stored numerical random keys and pinned source
+        config recorded a traceable 100-ID Gaussian-mixture result; a separate validator recomputed
+        raw-record aggregates, paired intervals, and source hashes.
+      </p>
+
+      <section class="warning" aria-labelledby="boundary-title">
+        <h2 id="boundary-title">Interpretation boundary</h2>
+        <p>
+          This is not an exact reproduction, independent replication, complete Table 1 or Table 2
+          reproduction, publication score, or broad outperformance claim. It is a stored,
+          traceable comparison against the multinomial baseline only. This static package
+          does not verify a source checkout or numerical execution, and it does not establish
+          source-hash identity, execution provenance, paper-level results, or significance.
+        </p>
+      </section>
+
+      <section aria-labelledby="record-title">
+        <h2 id="record-title">Recorded configuration</h2>
+        <div class="card">
+          <ul>
+            <li>Released <code>zgbkdlm/diffres</code> source at <code>767effe3e755067eb8a04422597fbf37eb8ab754</code>.</li>
+            <li>Stored Monte Carlo IDs 0–99; 10,000 particles; 1,000 sliced-Wasserstein projections.</li>
+            <li>Probability-flow ODE, <code>T=3</code>, 128 steps, Jentzen--Kloeden integrator.</li>
+            <li>Diffusion resampling and multinomial baseline; no other Table 1 baselines.</li>
+          </ul>
+        </div>
+      </section>
+
+      <section aria-labelledby="results-title">
+        <h2 id="results-title">Historical stored aggregates</h2>
+        <table>
+          <thead>
+            <tr>
+              <th scope="col">Metric (lower is better)</th>
+              <th scope="col" class="number">Diffusion resampling</th>
+              <th scope="col" class="number">Multinomial baseline</th>
+            </tr>
+          </thead>
+          <tbody>
+            <tr>
+              <td>Sliced Wasserstein L1, mean ± population SD</td>
+              <td class="number">0.0807686 ± 0.0212899</td>
+              <td class="number">0.0824334 ± 0.0250323</td>
+            </tr>
+            <tr>
+              <td>Squared posterior-mean residual L2, mean ± population SD</td>
+              <td class="number">0.0373908 ± 0.0299358</td>
+              <td class="number">0.0378330 ± 0.0442877</td>
+            </tr>
+          </tbody>
+        </table>
+        <p>
+          The recorded paired 95% t intervals cross zero for both metrics. They are not evidence of
+          a statistically significant improvement.
+        </p>
+      </section>
+
+      <section aria-labelledby="evidence-title">
+        <h2 id="evidence-title">Inspect the evidence</h2>
+        <ul>
+          <li><a href="evidence/gms-selected-ids-0-99.json">Raw 100-ID result</a></li>
+          <li><a href="evidence/gms-selected-ids-0-99-validation.json">Historic validation record</a></li>
+          <li><a href="evidence/validate_gms_selected_result.py">Validator source</a></li>
+          <li><a href="evidence/verify_stored_gms_evidence.py">Self-contained stored-evidence verifier</a></li>
+          <li><a href="evidence/README.md">Evidence guide and validator limits</a></li>
+          <li><a href="MANIFEST.md">Source and artefact manifest</a></li>
+          <li><a href="README.md">Full README, safety note, and licensing</a></li>
+        </ul>
+      </section>
+
+      <section aria-labelledby="safety-title">
+        <h2 id="safety-title">Selection and safety</h2>
+        <p>
+          The package contains only static presentation files, two historical JSON records, a
+          non-networked audit script, and licensing/provenance notes. It deliberately excludes the
+          upstream source tree, virtual environments, caches, runners, model artefacts, partial
+          studies, and LGSSM material.
+        </p>
+      </section>
+
+      <footer>
+        <a href="https://arxiv.org/abs/2512.10401">Paper</a> ·
+        <a href="https://openreview.net/forum?id=M0e5XORjAW">OpenReview</a> ·
+        <a href="https://github.com/zgbkdlm/diffres">released source</a> ·
+        <a href="NOTICE">attribution and licence notice</a>
+      </footer>
+    </main>
+  </body>
 </html>