{"id":"48b07091b45342847bc9f86b8b4db215aa4ddad626c892d6a4ff1f2de0626fdf","title":"Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers","argument":"The authors of “Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers” argue that a reader can give consistent outputs for equivalent neural-network weights without thereby showing that it understands the represented content. Hidden-neuron permutations and positive ReLU rescalings change parameter coordinates while preserving the represented function, giving the authors a way to test whether readers rely on arbitrary coordinates or functional information.\n\nTheir evaluation compares a statistics reader and a normalized set reader on shape recognition, and tests frozen language models and a reasoning setup on weight-reading prompts. The abstract reports that the normalized set reader retains accuracy with unchanged predictions under joint transformations; it also reports disagreement between fresh, identical raw prompts. Canonicalization makes equivalent prompts identical, but the authors explicitly caution that this neither establishes recognition nor guarantees identical generated responses.\n\nThe authors present joint correctness/agreement metrics, semantic-cue controls, fresh repeats and exact-rational rounding bounds as an inspectable evaluation protocol. They restrict their claims to the tested fields, prompts and execution configurations. The abstract’s accuracy values appear as unreplaced placeholders, so their numerical results cannot be read from","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied abstract contains unreplaced placeholders for the reported accuracy and disagreement rates, so those numerical results cannot be assessed from this evidence. The evidence is the arXiv abstract and page, not the full paper; detailed methods and results beyond the abstract are unavailable here.","speaker":{"name":"Cha, Sungguk, Kim, Mintae, Han, Youngsub, Jeon, Byoung-Ki, Lee, Sangyeob","roleAtSource":"Article author"},"source":{"title":"Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.05798","publishedAt":"2026-11-04","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors test whether visual weight readers separate functional information from arbitrary parameter coordinates using transformations that preserve the represented function.","excerpt":"We audit this separation with hidden-neuron permutations and positive ReLU rescalings that preserve the represented function.","url":"https://arxiv.org/abs/2611.05798","relationship":"Source-stated"},{"text":"For shape recognition, the abstract reports that a normalized set reader retains accuracy under joint transformations, with unchanged predictions; the reported accuracy value is unavailable in the supplied text.","excerpt":"a normalized set reader retains \\ShapeNormSet\\% with unchanged predictions.","url":"https://arxiv.org/abs/2611.05798","relationship":"Source-stated"},{"text":"Fresh identical raw prompts disagree on some origins in the reasoning experiment; the abstract’s reported rate is an unreplaced placeholder.","excerpt":"fresh identical raw prompts disagree on \\FrontRawRepeatRange\\% of origins.","url":"https://arxiv.org/abs/2611.05798","relationship":"Source-stated"},{"text":"Canonicalization makes equivalent prompts identical, but the authors say this does not establish recognition or guarantee identical generated responses.","excerpt":"Canonicalization makes equivalent prompts identical, but this neither establishes recognition nor guarantees identical generated responses.","url":"https://arxiv.org/abs/2611.05798","relationship":"Source-stated"},{"text":"The authors describe the work as an inspectable evaluation protocol and limit its claims to the tested fields, prompts and execution configurations.","excerpt":"The result is an inspectable evaluation protocol for visual weight readers, with claims restricted to the tested fields, prompts, and execution configurations.","url":"https://arxiv.org/abs/2611.05798","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.05798","checkedAt":"2026-11-09T12:50:07.880Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How well do the reported results and evaluation protocol generalize beyond the tested fields, prompts and execution configurations?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers","summary":"The authors of “Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers” argue that a reader can give consistent outputs for equivalent neural-network weights without thereby showing that it understands the represented content. Hidden-neuron permutations and positive ReLU rescalings change parameter coordinates while preserving the represented function, giving the authors a way to test whether readers rely on arbitrary coordinates or functional information.\n\nTheir evaluation compares a statistics reader and a normalized set reader on shape recognition, and tests frozen language models and a reasoning setup on weight-reading prompts. The abstract reports that the normalized set reader retains accuracy with unchanged predictions under joint transformations; it also reports disagreement between fresh, identical raw prompts. Canonicalization makes equivalent prompts identical, but the authors explicitly caution that this neither establishes recognition nor guarantees identical generated responses.\n\nThe authors present joint correctness/agreement metrics, semantic-cue controls, fresh repeats and exact-rational rounding bounds as an inspectable evaluation protocol. They restrict their claims to the tested fields, prompts and execution configurations. The abstract’s accuracy values appear as unreplaced placeholders, so their numerical results cannot be read from","question":"How well do the reported results and evaluation protocol generalize beyond the tested fields, prompts and execution configurations?","claims":[{"text":"The authors test whether visual weight readers separate functional information from arbitrary parameter coordinates using transformations that preserve the represented function.","excerpt":"We audit this separation with hidden-neuron permutations and positive ReLU rescalings that preserve the represented function.","support":"source-stated"},{"text":"For shape recognition, the abstract reports that a normalized set reader retains accuracy under joint transformations, with unchanged predictions; the reported accuracy value is unavailable in the supplied text.","excerpt":"a normalized set reader retains \\ShapeNormSet\\% with unchanged predictions.","support":"source-stated"},{"text":"Fresh identical raw prompts disagree on some origins in the reasoning experiment; the abstract’s reported rate is an unreplaced placeholder.","excerpt":"fresh identical raw prompts disagree on \\FrontRawRepeatRange\\% of origins.","support":"source-stated"},{"text":"Canonicalization makes equivalent prompts identical, but the authors say this does not establish recognition or guarantee identical generated responses.","excerpt":"Canonicalization makes equivalent prompts identical, but this neither establishes recognition nor guarantees identical generated responses.","support":"source-stated"},{"text":"The authors describe the work as an inspectable evaluation protocol and limit its claims to the tested fields, prompts and execution configurations.","excerpt":"The result is an inspectable evaluation protocol for visual weight readers, with claims restricted to the tested fields, prompts, and execution configurations.","support":"source-stated"}],"limitations":["The supplied abstract contains unreplaced placeholders for the reported accuracy and disagreement rates, so those numerical results cannot be assessed from this evidence.","The evidence is the arXiv abstract and page, not the full paper; detailed methods and results beyond the abstract are unavailable here."],"mediaId":"61b7f07a03db13f7965d4e2a7e5148abcb6f601b25fe5b271ad30a1bc5989da9","correctionReason":null,"id":"48b07091b45342847bc9f86b8b4db215aa4ddad626c892d6a4ff1f2de0626fdf","publicationId":"8ce9055a-86a5-4996-81af-6b875c5c57a2","version":1,"sourceUrl":"https://arxiv.org/abs/2611.05798","sourceTitle":"Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-04","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-05T05:00:28.404Z","createdAt":"2026-11-09T12:19:55.897Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"61b7f07a03db13f7965d4e2a7e5148abcb6f601b25fe5b271ad30a1bc5989da9","kind":"article","url":"https://arxiv.org/abs/2611.05798","sourceUrl":"https://arxiv.org/abs/2611.05798","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-09T12:19:01.990Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.05798","checkedAt":"2026-11-09T12:50:07.880Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"61b7f07a03db13f7965d4e2a7e5148abcb6f601b25fe5b271ad30a1bc5989da9","confidence":0.64,"provider":"typesafe","model":"jev-1.13.0","inputHash":"e431dd29888510f3779333a67251763fc814a38cd559e97cbca8d5400d05c6b6","decidedAt":"2026-11-09T12:19:03.808Z"},"provenance":{"textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observationId":"1749edcd6352f6d939512d2ddc7f222afd064d3efee427babb8c4c4227a769db","retrievedAt":"2026-11-09T12:19:01.687Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"c759668c-6ffc-472b-8d15-f25f6daf1568"},"reviewer":{"model":"gpt-6-luna","runId":"36c283c8-9d27-4381-8aeb-4e662562cffe"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-09T12:19:55.897Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Cha, Sungguk","url":null,"kind":"unknown"},{"name":"Kim, Mintae","url":null,"kind":"unknown"},{"name":"Han, Youngsub","url":null,"kind":"unknown"},{"name":"Jeon, Byoung-Ki","url":null,"kind":"unknown"},{"name":"Lee, Sangyeob","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Cha, Sungguk","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Cha, Sungguk","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05798","locator":"head/meta[name=author|citation_author]"},"assertionId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T12:50:07.880Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c","sourceUrl":"https://arxiv.org/abs/2611.05798","contentSha256":"9d71434f4fdacf0e480126541608331915c3ecdd352dc1749c020132a54ab6d7","textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observedAt":"2026-11-09T12:50:07.880Z"}},{"actor":{"name":"Kim, Mintae","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kim, Mintae","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05798","locator":"head/meta[name=author|citation_author]"},"assertionId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T12:50:07.880Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c","sourceUrl":"https://arxiv.org/abs/2611.05798","contentSha256":"9d71434f4fdacf0e480126541608331915c3ecdd352dc1749c020132a54ab6d7","textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observedAt":"2026-11-09T12:50:07.880Z"}},{"actor":{"name":"Han, Youngsub","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Han, Youngsub","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05798","locator":"head/meta[name=author|citation_author]"},"assertionId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T12:50:07.880Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c","sourceUrl":"https://arxiv.org/abs/2611.05798","contentSha256":"9d71434f4fdacf0e480126541608331915c3ecdd352dc1749c020132a54ab6d7","textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observedAt":"2026-11-09T12:50:07.880Z"}},{"actor":{"name":"Jeon, Byoung-Ki","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Jeon, Byoung-Ki","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05798","locator":"head/meta[name=author|citation_author]"},"assertionId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T12:50:07.880Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c","sourceUrl":"https://arxiv.org/abs/2611.05798","contentSha256":"9d71434f4fdacf0e480126541608331915c3ecdd352dc1749c020132a54ab6d7","textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observedAt":"2026-11-09T12:50:07.880Z"}},{"actor":{"name":"Lee, Sangyeob","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Lee, Sangyeob","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05798","locator":"head/meta[name=author|citation_author]"},"assertionId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T12:50:07.880Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c","sourceUrl":"https://arxiv.org/abs/2611.05798","contentSha256":"9d71434f4fdacf0e480126541608331915c3ecdd352dc1749c020132a54ab6d7","textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observedAt":"2026-11-09T12:50:07.880Z"}}],"participationCoverage":{"total":5,"returned":5,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.05798","snapshotId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c","contentSha256":"9d71434f4fdacf0e480126541608331915c3ecdd352dc1749c020132a54ab6d7","textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observedAt":"2026-11-09T12:50:07.880Z","basis":"publisher-page-metadata","sourceLinks":43,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/48b07091b45342847bc9f86b8b4db215aa4ddad626c892d6a4ff1f2de0626fdf/source"}},"canonicalUrl":"https://escapemeridian.com/readings/consistency-is-not-understanding-orbit-audits-for-neural-field-48b07091","formats":{"html":"/readings/consistency-is-not-understanding-orbit-audits-for-neural-field-48b07091","markdown":"/readings/consistency-is-not-understanding-orbit-audits-for-neural-field-48b07091.md","json":"/readings/consistency-is-not-understanding-orbit-audits-for-neural-field-48b07091.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[{"id":"term-neural-network","name":"Neural network","explanation":"A system of connected simple calculations, loosely inspired by the brain, that learns patterns from data.","referenceUrl":"https://developers.google.com/machine-learning/glossary#neural-network"}],"entities":[{"id":"arxiv","name":"arXiv","kind":"organization","context":"arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.","referenceUrls":["https://arxiv.org/","https://info.arxiv.org/help/api/index.html"],"path":"/entities/arxiv","json":"/entities/arxiv.json","markdown":"/entities/arxiv.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"48b07091b45342847bc9f86b8b4db215aa4ddad626c892d6a4ff1f2de0626fdf","revision":"8ce9055a-86a5-4996-81af-6b875c5c57a2","revisionKind":"publication-id","title":"Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers","url":"https://escapemeridian.com/readings/consistency-is-not-understanding-orbit-audits-for-neural-field-48b07091"},"source":{"type":"source","id":"https://arxiv.org/abs/2611.05798","revision":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","revisionKind":"sha256-source-text","title":"Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers","url":"https://arxiv.org/abs/2611.05798"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter."},"revisions":[{"id":"48b07091b45342847bc9f86b8b4db215aa4ddad626c892d6a4ff1f2de0626fdf","title":"Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers","argument":"The authors of “Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers” argue that a reader can give consistent outputs for equivalent neural-network weights without thereby showing that it understands the represented content. Hidden-neuron permutations and positive ReLU rescalings change parameter coordinates while preserving the represented function, giving the authors a way to test whether readers rely on arbitrary coordinates or functional information.\n\nTheir evaluation compares a statistics reader and a normalized set reader on shape recognition, and tests frozen language models and a reasoning setup on weight-reading prompts. The abstract reports that the normalized set reader retains accuracy with unchanged predictions under joint transformations; it also reports disagreement between fresh, identical raw prompts. Canonicalization makes equivalent prompts identical, but the authors explicitly caution that this neither establishes recognition nor guarantees identical generated responses.\n\nThe authors present joint correctness/agreement metrics, semantic-cue controls, fresh repeats and exact-rational rounding bounds as an inspectable evaluation protocol. They restrict their claims to the tested fields, prompts and execution configurations. The abstract’s accuracy values appear as unreplaced placeholders, so their numerical results cannot be read from","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied abstract contains unreplaced placeholders for the reported accuracy and disagreement rates, so those numerical results cannot be assessed from this evidence. The evidence is the arXiv abstract and page, not the full paper; detailed methods and results beyond the abstract are unavailable here.","speaker":{"name":"Cha, Sungguk, Kim, Mintae, Han, Youngsub, Jeon, Byoung-Ki, Lee, Sangyeob","roleAtSource":"Article author"},"source":{"title":"Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers","publisher":"arXiv cs.AI","format":"Article","url":"https://arxiv.org/abs/2611.05798","publishedAt":"2026-11-04","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors test whether visual weight readers separate functional information from arbitrary parameter coordinates using transformations that preserve the represented function.","excerpt":"We audit this separation with hidden-neuron permutations and positive ReLU rescalings that preserve the represented function.","url":"https://arxiv.org/abs/2611.05798","relationship":"Source-stated"},{"text":"For shape recognition, the abstract reports that a normalized set reader retains accuracy under joint transformations, with unchanged predictions; the reported accuracy value is unavailable in the supplied text.","excerpt":"a normalized set reader retains \\ShapeNormSet\\% with unchanged predictions.","url":"https://arxiv.org/abs/2611.05798","relationship":"Source-stated"},{"text":"Fresh identical raw prompts disagree on some origins in the reasoning experiment; the abstract’s reported rate is an unreplaced placeholder.","excerpt":"fresh identical raw prompts disagree on \\FrontRawRepeatRange\\% of origins.","url":"https://arxiv.org/abs/2611.05798","relationship":"Source-stated"},{"text":"Canonicalization makes equivalent prompts identical, but the authors say this does not establish recognition or guarantee identical generated responses.","excerpt":"Canonicalization makes equivalent prompts identical, but this neither establishes recognition nor guarantees identical generated responses.","url":"https://arxiv.org/abs/2611.05798","relationship":"Source-stated"},{"text":"The authors describe the work as an inspectable evaluation protocol and limit its claims to the tested fields, prompts and execution configurations.","excerpt":"The result is an inspectable evaluation protocol for visual weight readers, with claims restricted to the tested fields, prompts, and execution configurations.","url":"https://arxiv.org/abs/2611.05798","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.05798","checkedAt":"2026-11-09T12:50:07.880Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How well do the reported results and evaluation protocol generalize beyond the tested fields, prompts and execution configurations?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers","summary":"The authors of “Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers” argue that a reader can give consistent outputs for equivalent neural-network weights without thereby showing that it understands the represented content. Hidden-neuron permutations and positive ReLU rescalings change parameter coordinates while preserving the represented function, giving the authors a way to test whether readers rely on arbitrary coordinates or functional information.\n\nTheir evaluation compares a statistics reader and a normalized set reader on shape recognition, and tests frozen language models and a reasoning setup on weight-reading prompts. The abstract reports that the normalized set reader retains accuracy with unchanged predictions under joint transformations; it also reports disagreement between fresh, identical raw prompts. Canonicalization makes equivalent prompts identical, but the authors explicitly caution that this neither establishes recognition nor guarantees identical generated responses.\n\nThe authors present joint correctness/agreement metrics, semantic-cue controls, fresh repeats and exact-rational rounding bounds as an inspectable evaluation protocol. They restrict their claims to the tested fields, prompts and execution configurations. The abstract’s accuracy values appear as unreplaced placeholders, so their numerical results cannot be read from","question":"How well do the reported results and evaluation protocol generalize beyond the tested fields, prompts and execution configurations?","claims":[{"text":"The authors test whether visual weight readers separate functional information from arbitrary parameter coordinates using transformations that preserve the represented function.","excerpt":"We audit this separation with hidden-neuron permutations and positive ReLU rescalings that preserve the represented function.","support":"source-stated"},{"text":"For shape recognition, the abstract reports that a normalized set reader retains accuracy under joint transformations, with unchanged predictions; the reported accuracy value is unavailable in the supplied text.","excerpt":"a normalized set reader retains \\ShapeNormSet\\% with unchanged predictions.","support":"source-stated"},{"text":"Fresh identical raw prompts disagree on some origins in the reasoning experiment; the abstract’s reported rate is an unreplaced placeholder.","excerpt":"fresh identical raw prompts disagree on \\FrontRawRepeatRange\\% of origins.","support":"source-stated"},{"text":"Canonicalization makes equivalent prompts identical, but the authors say this does not establish recognition or guarantee identical generated responses.","excerpt":"Canonicalization makes equivalent prompts identical, but this neither establishes recognition nor guarantees identical generated responses.","support":"source-stated"},{"text":"The authors describe the work as an inspectable evaluation protocol and limit its claims to the tested fields, prompts and execution configurations.","excerpt":"The result is an inspectable evaluation protocol for visual weight readers, with claims restricted to the tested fields, prompts, and execution configurations.","support":"source-stated"}],"limitations":["The supplied abstract contains unreplaced placeholders for the reported accuracy and disagreement rates, so those numerical results cannot be assessed from this evidence.","The evidence is the arXiv abstract and page, not the full paper; detailed methods and results beyond the abstract are unavailable here."],"mediaId":"61b7f07a03db13f7965d4e2a7e5148abcb6f601b25fe5b271ad30a1bc5989da9","correctionReason":null,"id":"48b07091b45342847bc9f86b8b4db215aa4ddad626c892d6a4ff1f2de0626fdf","publicationId":"8ce9055a-86a5-4996-81af-6b875c5c57a2","version":1,"sourceUrl":"https://arxiv.org/abs/2611.05798","sourceTitle":"Consistency Is Not Understanding: Orbit Audits for Neural Field Weight Readers","publisher":"arXiv cs.AI","publishedAt":null,"originalPublished":{"value":"2026-11-04","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-05T05:00:28.404Z","createdAt":"2026-11-09T12:19:55.897Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"61b7f07a03db13f7965d4e2a7e5148abcb6f601b25fe5b271ad30a1bc5989da9","kind":"article","url":"https://arxiv.org/abs/2611.05798","sourceUrl":"https://arxiv.org/abs/2611.05798","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-09T12:19:01.990Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2611.05798","checkedAt":"2026-11-09T12:50:07.880Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"61b7f07a03db13f7965d4e2a7e5148abcb6f601b25fe5b271ad30a1bc5989da9","confidence":0.64,"provider":"typesafe","model":"jev-1.13.0","inputHash":"e431dd29888510f3779333a67251763fc814a38cd559e97cbca8d5400d05c6b6","decidedAt":"2026-11-09T12:19:03.808Z"},"provenance":{"textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observationId":"1749edcd6352f6d939512d2ddc7f222afd064d3efee427babb8c4c4227a769db","retrievedAt":"2026-11-09T12:19:01.687Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"c759668c-6ffc-472b-8d15-f25f6daf1568"},"reviewer":{"model":"gpt-6-luna","runId":"36c283c8-9d27-4381-8aeb-4e662562cffe"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-09T12:19:55.897Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Cha, Sungguk","url":null,"kind":"unknown"},{"name":"Kim, Mintae","url":null,"kind":"unknown"},{"name":"Han, Youngsub","url":null,"kind":"unknown"},{"name":"Jeon, Byoung-Ki","url":null,"kind":"unknown"},{"name":"Lee, Sangyeob","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Cha, Sungguk","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Cha, Sungguk","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05798","locator":"head/meta[name=author|citation_author]"},"assertionId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T12:50:07.880Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c","sourceUrl":"https://arxiv.org/abs/2611.05798","contentSha256":"9d71434f4fdacf0e480126541608331915c3ecdd352dc1749c020132a54ab6d7","textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observedAt":"2026-11-09T12:50:07.880Z"}},{"actor":{"name":"Kim, Mintae","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Kim, Mintae","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05798","locator":"head/meta[name=author|citation_author]"},"assertionId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T12:50:07.880Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c","sourceUrl":"https://arxiv.org/abs/2611.05798","contentSha256":"9d71434f4fdacf0e480126541608331915c3ecdd352dc1749c020132a54ab6d7","textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observedAt":"2026-11-09T12:50:07.880Z"}},{"actor":{"name":"Han, Youngsub","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Han, Youngsub","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05798","locator":"head/meta[name=author|citation_author]"},"assertionId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T12:50:07.880Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c","sourceUrl":"https://arxiv.org/abs/2611.05798","contentSha256":"9d71434f4fdacf0e480126541608331915c3ecdd352dc1749c020132a54ab6d7","textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observedAt":"2026-11-09T12:50:07.880Z"}},{"actor":{"name":"Jeon, Byoung-Ki","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Jeon, Byoung-Ki","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05798","locator":"head/meta[name=author|citation_author]"},"assertionId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T12:50:07.880Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c","sourceUrl":"https://arxiv.org/abs/2611.05798","contentSha256":"9d71434f4fdacf0e480126541608331915c3ecdd352dc1749c020132a54ab6d7","textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observedAt":"2026-11-09T12:50:07.880Z"}},{"actor":{"name":"Lee, Sangyeob","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Lee, Sangyeob","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2611.05798","locator":"head/meta[name=author|citation_author]"},"assertionId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-09T12:50:07.880Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c","sourceUrl":"https://arxiv.org/abs/2611.05798","contentSha256":"9d71434f4fdacf0e480126541608331915c3ecdd352dc1749c020132a54ab6d7","textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observedAt":"2026-11-09T12:50:07.880Z"}}],"participationCoverage":{"total":5,"returned":5,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2611.05798","snapshotId":"a607cb0f768b366372a5074d33a33b8401c7ed1986565083f2213b8b7f3ed89c","contentSha256":"9d71434f4fdacf0e480126541608331915c3ecdd352dc1749c020132a54ab6d7","textHash":"28318f65936b72c71a7eda14ad4d07002b333eb083b3802b10fca61d0d82e4f3","observedAt":"2026-11-09T12:50:07.880Z","basis":"publisher-page-metadata","sourceLinks":43,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/48b07091b45342847bc9f86b8b4db215aa4ddad626c892d6a4ff1f2de0626fdf/source"}}}],"audits":[{"publicationId":"8ce9055a-86a5-4996-81af-6b875c5c57a2","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading attributes the argument and results to the paper's authors, using the arXiv title and abstract. It does not present the authors' claims as established fact.","excerpt":"Authors:Sungguk Cha, Mintae Kim, Youngsub Han, Byoung-Ki Jeon, Sangyeob Lee"},{"dimension":"summary","status":"uncertain","explanation":"The summary is cut off mid-sentence, ending with 'cannot be read from'. Smallest repair: complete it, for example by adding 'the supplied abstract.' Central terms are not explained for a general reader, including neural field, weight reader, ReLU, hidden-neuron permutations and canonicalization. Without them, a non-specialist cannot follow what was tested or why it matters. Keep the exact terms but explain them through term-library tooltips, adding entries where missing. Apart from these gaps, the faithful account and the authors' caution are preserved.","excerpt":null},{"dimension":"date","status":"pass","explanation":"The originalPublished date of 2026-11-04 matches the v1 submission date shown on the arXiv page. The 2026-11-05 feed date is correctly not used.","excerpt":"[Submitted on 4 Nov 2026]"},{"dimension":"support","status":"pass","explanation":"Each claim matches the abstract and is flagged as source-stated. The reading correctly notes that the numeric results are unreplaced placeholders. The question about generalization follows from the authors' own restriction of scope. No benefits, costs or outcomes are invented.","excerpt":"Canonicalization makes equivalent prompts identical, but this neither establishes recognition nor guarantees identical generated responses."},{"dimension":"media","status":"pass","explanation":"The selected media is the verified arXiv abstract page, which is the direct original source.","excerpt":null}],"createdAt":"2026-11-09T12:19:55.897Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}