{"id":"bebc6ea4b57286304d6ed2e683162cff91e6fc023ccadfb2c5879cf36354d794","title":"arXiv q-bio.NC: Heng Zhang introduces RECAP for reservoir dynamics","argument":"Heng Zhang introduces RECAP, a method for image classification that pairs untrained reservoir dynamics with a self-organizing Hebbian prototype readout. Instead of training through error backpropagation, RECAP turns time-averaged reservoir responses into activation levels, records which reservoir units activate together, and updates class-wise prototype matrices using a potentiation-decay rule. At inference, it matches new responses against those prototypes by overlap. Zhang presents this design as avoiding error backpropagation and as compatible with online prototype updates, where prototypes can be updated incrementally. The paper reports robustness behavior on MNIST-C: RECAP remains robust under diverse corruptions despite not being exposed to corrupted training samples. This makes the work relevant to whether a learning method built around local updates can handle image corruption rather than relying on end-to-end gradient optimization. The supplied abstract does not give quantitative results or enough detail to judge performance across corruptions; its claim is limited to the reported MNIST-C experiments.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied abstract reports robustness on MNIST-C but provides no quantitative results or corruption-by-corruption detail, so the extent of the reported performance cannot be assessed from this evidence alone.","speaker":{"name":"Zhang, Heng","roleAtSource":"Article author"},"source":{"title":"RECAP: Local Hebbian Prototype Learning as a Self-Organizing Readout for Reservoir Dynamics","publisher":"arXiv q-bio.NC","format":"Article","url":"https://arxiv.org/abs/2603.06639","publishedAt":"2026-02-25","dateBasis":"Date basis not provided."},"evidence":[{"text":"Heng Zhang introduces RECAP, a method for image classification that combines untrained reservoir dynamics with a self-organizing Hebbian prototype readout.","excerpt":"We introduce RECAP (Reservoir Computing with Hebbian Co-Activation Prototypes), a bio-inspired learning strategy for robust image classification that couples untrained reservoir dynamics with a self-organizing Hebbian prototype readout.","url":"https://arxiv.org/abs/2603.06639","relationship":"Source-stated"},{"text":"RECAP forms activation levels from time-averaged reservoir responses, records co-activation among unit pairs, and incrementally updates class-wise prototype matrices with a potentiation-decay rule.","excerpt":"RECAP discretizes time-averaged reservoir responses into activation levels, constructs a co-activation mask over reservoir unit pairs, and incrementally updates class-wise prototype matrices via a Hebbian-like potentiation-decay rule.","url":"https://arxiv.org/abs/2603.06639","relationship":"Source-stated"},{"text":"Zhang reports robustness behavior on MNIST-C under diverse corruptions, without exposure to corrupted training samples.","excerpt":"We illustrate the resulting robustness behavior on MNIST-C, where RECAP remains robust under diverse corruptions without exposure to corrupted training samples.","url":"https://arxiv.org/abs/2603.06639","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2603.06639","checkedAt":"2026-11-12T00:04:08.110Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does RECAP’s performance vary across the corruptions tested on MNIST-C, and what quantitative results support the reported robustness?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv q-bio.NC: Heng Zhang introduces RECAP for reservoir dynamics","summary":"Heng Zhang introduces RECAP, a method for image classification that pairs untrained reservoir dynamics with a self-organizing Hebbian prototype readout. Instead of training through error backpropagation, RECAP turns time-averaged reservoir responses into activation levels, records which reservoir units activate together, and updates class-wise prototype matrices using a potentiation-decay rule. At inference, it matches new responses against those prototypes by overlap. Zhang presents this design as avoiding error backpropagation and as compatible with online prototype updates, where prototypes can be updated incrementally. The paper reports robustness behavior on MNIST-C: RECAP remains robust under diverse corruptions despite not being exposed to corrupted training samples. This makes the work relevant to whether a learning method built around local updates can handle image corruption rather than relying on end-to-end gradient optimization. The supplied abstract does not give quantitative results or enough detail to judge performance across corruptions; its claim is limited to the reported MNIST-C experiments.","question":"How does RECAP’s performance vary across the corruptions tested on MNIST-C, and what quantitative results support the reported robustness?","claims":[{"text":"Heng Zhang introduces RECAP, a method for image classification that combines untrained reservoir dynamics with a self-organizing Hebbian prototype readout.","excerpt":"We introduce RECAP (Reservoir Computing with Hebbian Co-Activation Prototypes), a bio-inspired learning strategy for robust image classification that couples untrained reservoir dynamics with a self-organizing Hebbian prototype readout.","support":"source-stated"},{"text":"RECAP forms activation levels from time-averaged reservoir responses, records co-activation among unit pairs, and incrementally updates class-wise prototype matrices with a potentiation-decay rule.","excerpt":"RECAP discretizes time-averaged reservoir responses into activation levels, constructs a co-activation mask over reservoir unit pairs, and incrementally updates class-wise prototype matrices via a Hebbian-like potentiation-decay rule.","support":"source-stated"},{"text":"Zhang reports robustness behavior on MNIST-C under diverse corruptions, without exposure to corrupted training samples.","excerpt":"We illustrate the resulting robustness behavior on MNIST-C, where RECAP remains robust under diverse corruptions without exposure to corrupted training samples.","support":"source-stated"}],"limitations":["The supplied abstract reports robustness on MNIST-C but provides no quantitative results or corruption-by-corruption detail, so the extent of the reported performance cannot be assessed from this evidence alone."],"mediaId":"fb141e7ac0f257bad16d542c19f701b17dd35c1369c9b97c8a318bfe1f48882e","correctionReason":null,"id":"bebc6ea4b57286304d6ed2e683162cff91e6fc023ccadfb2c5879cf36354d794","publicationId":"556f5d80-bf74-4761-8756-c4b8f54b6e2f","version":1,"sourceUrl":"https://arxiv.org/abs/2603.06639","sourceTitle":"RECAP: Local Hebbian Prototype Learning as a Self-Organizing Readout for Reservoir Dynamics","publisher":"arXiv q-bio.NC","publishedAt":null,"originalPublished":{"value":"2026-02-25","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-11T05:01:34.530Z","createdAt":"2026-11-11T23:33:46.896Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"fb141e7ac0f257bad16d542c19f701b17dd35c1369c9b97c8a318bfe1f48882e","kind":"article","url":"https://arxiv.org/abs/2603.06639","sourceUrl":"https://arxiv.org/abs/2603.06639","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-11T23:32:41.027Z","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/2603.06639","checkedAt":"2026-11-12T00:04:08.110Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"fb141e7ac0f257bad16d542c19f701b17dd35c1369c9b97c8a318bfe1f48882e","confidence":0.63,"provider":"typesafe","model":"jev-1.13.0","inputHash":"48c32e223507ab51c6c33cdbabcb6f31c39689914500504bd69e75964cfd04f8","decidedAt":"2026-11-11T23:32:42.948Z"},"provenance":{"textHash":"2352a395b2c72a579019a395d061968b35b6add61f3c153d523a9fa9f4f68a56","observationId":"7d56588f28dfe2181b8a408751458aad2343e6d68de090b56b9eba5f4d083900","retrievedAt":"2026-11-11T23:32:40.733Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"010814ed-aef5-4a02-8d9a-bfe9a1710ec1"},"reviewer":{"model":"gpt-6-luna","runId":"dddf112d-bab0-4235-8606-93fa610b3b5a"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-11T23:33:46.896Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Zhang, Heng","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Zhang, Heng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zhang, Heng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2603.06639","locator":"head/meta[name=author|citation_author]"},"assertionId":"5ee95646e7a23b7574a70e810c341a3a8c7cd4936e6926f4bc46dae1aead675d:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-12T00:04:08.110Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"5ee95646e7a23b7574a70e810c341a3a8c7cd4936e6926f4bc46dae1aead675d","sourceUrl":"https://arxiv.org/abs/2603.06639","contentSha256":"4eb39ffc77be82f8c286dd52b1a84528928717dd6db3a53e7feddb0b11e4168e","textHash":"2352a395b2c72a579019a395d061968b35b6add61f3c153d523a9fa9f4f68a56","observedAt":"2026-11-12T00:04:08.110Z"}}],"participationCoverage":{"total":1,"returned":1,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2603.06639","snapshotId":"5ee95646e7a23b7574a70e810c341a3a8c7cd4936e6926f4bc46dae1aead675d","contentSha256":"4eb39ffc77be82f8c286dd52b1a84528928717dd6db3a53e7feddb0b11e4168e","textHash":"2352a395b2c72a579019a395d061968b35b6add61f3c153d523a9fa9f4f68a56","observedAt":"2026-11-12T00:04:08.110Z","basis":"publisher-page-metadata","sourceLinks":46,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/bebc6ea4b57286304d6ed2e683162cff91e6fc023ccadfb2c5879cf36354d794/source"}},"canonicalUrl":"https://escapemeridian.com/readings/arxiv-q-bio-nc-heng-zhang-introduces-recap-for-reservoir-bebc6ea4","formats":{"html":"/readings/arxiv-q-bio-nc-heng-zhang-introduces-recap-for-reservoir-bebc6ea4","markdown":"/readings/arxiv-q-bio-nc-heng-zhang-introduces-recap-for-reservoir-bebc6ea4.md","json":"/readings/arxiv-q-bio-nc-heng-zhang-introduces-recap-for-reservoir-bebc6ea4.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":[],"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"},{"id":"time","name":"TIME","kind":"organization","context":"TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.","referenceUrls":["https://time.com/"],"path":"/entities/time","json":"/entities/time.json","markdown":"/entities/time.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"bebc6ea4b57286304d6ed2e683162cff91e6fc023ccadfb2c5879cf36354d794","revision":"556f5d80-bf74-4761-8756-c4b8f54b6e2f","revisionKind":"publication-id","title":"arXiv q-bio.NC: Heng Zhang introduces RECAP for reservoir dynamics","url":"https://escapemeridian.com/readings/arxiv-q-bio-nc-heng-zhang-introduces-recap-for-reservoir-bebc6ea4"},"source":{"type":"source","id":"https://arxiv.org/abs/2603.06639","revision":"2352a395b2c72a579019a395d061968b35b6add61f3c153d523a9fa9f4f68a56","revisionKind":"sha256-source-text","title":"RECAP: Local Hebbian Prototype Learning as a Self-Organizing Readout for Reservoir Dynamics","url":"https://arxiv.org/abs/2603.06639"},"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":"bebc6ea4b57286304d6ed2e683162cff91e6fc023ccadfb2c5879cf36354d794","title":"arXiv q-bio.NC: Heng Zhang introduces RECAP for reservoir dynamics","argument":"Heng Zhang introduces RECAP, a method for image classification that pairs untrained reservoir dynamics with a self-organizing Hebbian prototype readout. Instead of training through error backpropagation, RECAP turns time-averaged reservoir responses into activation levels, records which reservoir units activate together, and updates class-wise prototype matrices using a potentiation-decay rule. At inference, it matches new responses against those prototypes by overlap. Zhang presents this design as avoiding error backpropagation and as compatible with online prototype updates, where prototypes can be updated incrementally. The paper reports robustness behavior on MNIST-C: RECAP remains robust under diverse corruptions despite not being exposed to corrupted training samples. This makes the work relevant to whether a learning method built around local updates can handle image corruption rather than relying on end-to-end gradient optimization. The supplied abstract does not give quantitative results or enough detail to judge performance across corruptions; its claim is limited to the reported MNIST-C experiments.","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied abstract reports robustness on MNIST-C but provides no quantitative results or corruption-by-corruption detail, so the extent of the reported performance cannot be assessed from this evidence alone.","speaker":{"name":"Zhang, Heng","roleAtSource":"Article author"},"source":{"title":"RECAP: Local Hebbian Prototype Learning as a Self-Organizing Readout for Reservoir Dynamics","publisher":"arXiv q-bio.NC","format":"Article","url":"https://arxiv.org/abs/2603.06639","publishedAt":"2026-02-25","dateBasis":"Date basis not provided."},"evidence":[{"text":"Heng Zhang introduces RECAP, a method for image classification that combines untrained reservoir dynamics with a self-organizing Hebbian prototype readout.","excerpt":"We introduce RECAP (Reservoir Computing with Hebbian Co-Activation Prototypes), a bio-inspired learning strategy for robust image classification that couples untrained reservoir dynamics with a self-organizing Hebbian prototype readout.","url":"https://arxiv.org/abs/2603.06639","relationship":"Source-stated"},{"text":"RECAP forms activation levels from time-averaged reservoir responses, records co-activation among unit pairs, and incrementally updates class-wise prototype matrices with a potentiation-decay rule.","excerpt":"RECAP discretizes time-averaged reservoir responses into activation levels, constructs a co-activation mask over reservoir unit pairs, and incrementally updates class-wise prototype matrices via a Hebbian-like potentiation-decay rule.","url":"https://arxiv.org/abs/2603.06639","relationship":"Source-stated"},{"text":"Zhang reports robustness behavior on MNIST-C under diverse corruptions, without exposure to corrupted training samples.","excerpt":"We illustrate the resulting robustness behavior on MNIST-C, where RECAP remains robust under diverse corruptions without exposure to corrupted training samples.","url":"https://arxiv.org/abs/2603.06639","relationship":"Source-stated"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2603.06639","checkedAt":"2026-11-12T00:04:08.110Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does RECAP’s performance vary across the corruptions tested on MNIST-C, and what quantitative results support the reported robustness?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"arXiv q-bio.NC: Heng Zhang introduces RECAP for reservoir dynamics","summary":"Heng Zhang introduces RECAP, a method for image classification that pairs untrained reservoir dynamics with a self-organizing Hebbian prototype readout. Instead of training through error backpropagation, RECAP turns time-averaged reservoir responses into activation levels, records which reservoir units activate together, and updates class-wise prototype matrices using a potentiation-decay rule. At inference, it matches new responses against those prototypes by overlap. Zhang presents this design as avoiding error backpropagation and as compatible with online prototype updates, where prototypes can be updated incrementally. The paper reports robustness behavior on MNIST-C: RECAP remains robust under diverse corruptions despite not being exposed to corrupted training samples. This makes the work relevant to whether a learning method built around local updates can handle image corruption rather than relying on end-to-end gradient optimization. The supplied abstract does not give quantitative results or enough detail to judge performance across corruptions; its claim is limited to the reported MNIST-C experiments.","question":"How does RECAP’s performance vary across the corruptions tested on MNIST-C, and what quantitative results support the reported robustness?","claims":[{"text":"Heng Zhang introduces RECAP, a method for image classification that combines untrained reservoir dynamics with a self-organizing Hebbian prototype readout.","excerpt":"We introduce RECAP (Reservoir Computing with Hebbian Co-Activation Prototypes), a bio-inspired learning strategy for robust image classification that couples untrained reservoir dynamics with a self-organizing Hebbian prototype readout.","support":"source-stated"},{"text":"RECAP forms activation levels from time-averaged reservoir responses, records co-activation among unit pairs, and incrementally updates class-wise prototype matrices with a potentiation-decay rule.","excerpt":"RECAP discretizes time-averaged reservoir responses into activation levels, constructs a co-activation mask over reservoir unit pairs, and incrementally updates class-wise prototype matrices via a Hebbian-like potentiation-decay rule.","support":"source-stated"},{"text":"Zhang reports robustness behavior on MNIST-C under diverse corruptions, without exposure to corrupted training samples.","excerpt":"We illustrate the resulting robustness behavior on MNIST-C, where RECAP remains robust under diverse corruptions without exposure to corrupted training samples.","support":"source-stated"}],"limitations":["The supplied abstract reports robustness on MNIST-C but provides no quantitative results or corruption-by-corruption detail, so the extent of the reported performance cannot be assessed from this evidence alone."],"mediaId":"fb141e7ac0f257bad16d542c19f701b17dd35c1369c9b97c8a318bfe1f48882e","correctionReason":null,"id":"bebc6ea4b57286304d6ed2e683162cff91e6fc023ccadfb2c5879cf36354d794","publicationId":"556f5d80-bf74-4761-8756-c4b8f54b6e2f","version":1,"sourceUrl":"https://arxiv.org/abs/2603.06639","sourceTitle":"RECAP: Local Hebbian Prototype Learning as a Self-Organizing Readout for Reservoir Dynamics","publisher":"arXiv q-bio.NC","publishedAt":null,"originalPublished":{"value":"2026-02-25","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-11-11T05:01:34.530Z","createdAt":"2026-11-11T23:33:46.896Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"fb141e7ac0f257bad16d542c19f701b17dd35c1369c9b97c8a318bfe1f48882e","kind":"article","url":"https://arxiv.org/abs/2603.06639","sourceUrl":"https://arxiv.org/abs/2603.06639","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-11-11T23:32:41.027Z","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/2603.06639","checkedAt":"2026-11-12T00:04:08.110Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"fb141e7ac0f257bad16d542c19f701b17dd35c1369c9b97c8a318bfe1f48882e","confidence":0.63,"provider":"typesafe","model":"jev-1.13.0","inputHash":"48c32e223507ab51c6c33cdbabcb6f31c39689914500504bd69e75964cfd04f8","decidedAt":"2026-11-11T23:32:42.948Z"},"provenance":{"textHash":"2352a395b2c72a579019a395d061968b35b6add61f3c153d523a9fa9f4f68a56","observationId":"7d56588f28dfe2181b8a408751458aad2343e6d68de090b56b9eba5f4d083900","retrievedAt":"2026-11-11T23:32:40.733Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"010814ed-aef5-4a02-8d9a-bfe9a1710ec1"},"reviewer":{"model":"gpt-6-luna","runId":"dddf112d-bab0-4235-8606-93fa610b3b5a"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-11T23:33:46.896Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Zhang, Heng","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Zhang, Heng","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zhang, Heng","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2603.06639","locator":"head/meta[name=author|citation_author]"},"assertionId":"5ee95646e7a23b7574a70e810c341a3a8c7cd4936e6926f4bc46dae1aead675d:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-12T00:04:08.110Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"5ee95646e7a23b7574a70e810c341a3a8c7cd4936e6926f4bc46dae1aead675d","sourceUrl":"https://arxiv.org/abs/2603.06639","contentSha256":"4eb39ffc77be82f8c286dd52b1a84528928717dd6db3a53e7feddb0b11e4168e","textHash":"2352a395b2c72a579019a395d061968b35b6add61f3c153d523a9fa9f4f68a56","observedAt":"2026-11-12T00:04:08.110Z"}}],"participationCoverage":{"total":1,"returned":1,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2603.06639","snapshotId":"5ee95646e7a23b7574a70e810c341a3a8c7cd4936e6926f4bc46dae1aead675d","contentSha256":"4eb39ffc77be82f8c286dd52b1a84528928717dd6db3a53e7feddb0b11e4168e","textHash":"2352a395b2c72a579019a395d061968b35b6add61f3c153d523a9fa9f4f68a56","observedAt":"2026-11-12T00:04:08.110Z","basis":"publisher-page-metadata","sourceLinks":46,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/bebc6ea4b57286304d6ed2e683162cff91e6fc023ccadfb2c5879cf36354d794/source"}}}],"audits":[{"publicationId":"556f5d80-bf74-4761-8756-c4b8f54b6e2f","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The method and claims are correctly attributed to Heng Zhang, the sole listed author, on arXiv. Labelling the publisher as arXiv q-bio.NC is acceptable because the paper is cross-listed there, even though its primary category is cs.NE.","excerpt":"Authors:Heng Zhang"},{"dimension":"summary","status":"uncertain","explanation":"The mechanism is described accurately, but the reading drops the source's central motivation: brains rely on local plasticity, while backpropagation is not naturally aligned with local computation, so RECAP is offered as a bio-inspired alternative. Without that framing, a general reader cannot see why avoiding backpropagation matters. Central terms such as reservoir dynamics, Hebbian, backpropagation and MNIST-C also go unexplained. Smallest repair: add one sentence restoring the brain-inspired, local-plasticity rationale, and ensure tooltips cover these terms.","excerpt":"Robust perception in brains is often attributed to high-dimensional population activity together with local plasticity mechanisms that reinforce recurring structure."},{"dimension":"date","status":"pass","explanation":"The original publication date of 2026-02-25 matches the v1 submission date. The 10 Nov 2026 revision and the feed date are correctly not used as the original date.","excerpt":"[Submitted on 25 Feb 2026 (v1), last revised 10 Nov 2026 (this version, v2)]"},{"dimension":"support","status":"pass","explanation":"All claims match the abstract: untrained reservoir, Hebbian prototype readout, discretization, co-activation mask, potentiation-decay updates, overlap matching, no backpropagation, online updates, and MNIST-C robustness without corrupted training data. The limitation and the question correctly note that no quantitative results are given. No benefits are overstated.","excerpt":"We illustrate the resulting robustness behavior on MNIST-C, where RECAP remains robust under diverse corruptions without exposure to corrupted training samples."},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct link to the original arXiv abstract page, which is the most direct way for readers to access the source.","excerpt":null}],"createdAt":"2026-11-11T23:33:46.896Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}