{"schemaVersion":"maha-epistemic/1.0","evidencePolicyVersion":"mps/0.1","recordId":"urn:maha:record:mechanistic-interpretability-sae-encoder-decoder","canonicalPath":"/knowledge/mechanistic-interpretability/mechanisms/mechanistic-interpretability-sae-encoder-decoder","contentHash":"sha256:79445a65081305ad694085c50fbdd859d0a04b3503ec55f5839a3e76d5576ee1","generatedAt":"2026-08-31T04:42:05.329Z","publicationDecision":{"recordId":"urn:maha:record:mechanistic-interpretability-sae-encoder-decoder","publicEligible":true,"evaluatedAgainst":"maha-epistemic/1.0","reasons":[]},"claims":[{"id":"urn:maha:claim:mechanistic-interpretability-sae-encoder-decoder","scope":"Limited to Method, reconstruction and sparsity objectives, experiments, feature analysis, and limitations. in “Sparse Autoencoders Find Highly Interpretable Features in Language Models”; this candidate records the concept boundary and does not pool results from uncited systems or studies.","boundary":"SAE encoder decoder does not by itself establish system-level performance, safety, manufacturability, scalability, economic advantage, clinical benefit, or deployment readiness.","claimKind":"theoretical-model","sourceIds":["source-mechanistic-interpretability-sae"],"statement":"The cited source supports treating sae encoder decoder as a distinct mechanism within the stated mechanistic interpretability scope.","replication":{"asOfDate":"2026-08-24","assessment":"Independent replication and cross-platform transfer have not been compiled for this candidate; the evidence maturity refers only to the bounded source contract.","independentReplicationCount":null},"uncertainty":{"kind":"qualitative","statement":"No cross-source quantitative interval is asserted. Definitions, operating conditions, samples, instruments, and outcome measures must be checked against the exact cited locator during review."},"evidenceMaturity":"single-study"}],"sources":[{"id":"source-mechanistic-interpretability-sae","url":"https://arxiv.org/abs/2309.08600","title":"Sparse Autoencoders Find Highly Interpretable Features in Language Models","rights":{"note":"The candidate uses original boundary language and a short paraphrase linked to the cited source. No source passage, figure, or table is reproduced.","basis":"citation-with-paraphrase","quotationUsed":false},"authors":["Hoagy Cunningham","Aidan Ewart","Logan Riggs","Robert Huben","Lee Sharkey"],"boundary":"Sparse features and human labels do not establish completeness, unique decomposition, or causal faithfulness.","publisher":"arXiv","establishes":"The paper trains sparse autoencoders on language-model activations and evaluates specified reconstruction, sparsity, and interpretability properties.","identifiers":[{"value":"https://arxiv.org/abs/2309.08600","scheme":"url"}],"publishedAt":"2023-09-15","exactLocator":"Method, reconstruction and sparsity objectives, experiments, feature analysis, and limitations."}],"reviewEvents":[{"scope":"source-fidelity","verdict":"approve","reviewId":"epireview_2b020afefc17455785db3ae2a5a78f9b","rationale":"The encoder–decoder construction with reconstruction and sparsity objectives is named in the inspected method locator, so the mechanism is the paper’s own. The record claims no reconstruction loss, sparsity level, or feature count. This approval is limited to source-fidelity, record urn:maha:record:mechanistic-interpretability-sae-encoder-decoder, and exact target sha256:369c160f224159778ec08f586366d8c163e81afe67c769a0c5b9db269a01d54c; it is AI-assisted internal editorial review, not external expert review, peer review, consensus, independent reproduction, scientific validation, or commercial certification.","reviewedAt":"2026-08-31T04:41:56.303Z","reviewerId":"expert_maha-internal-editorial-v2","reviewMethod":"Record-specific exact-revision review against the inspected source location, bounded claim, non-claims, rights basis, and source identity. No external reviewer participated.","reviewerKind":"internal-editorial","reviewerRole":"AI-assisted internal source, scope, boundary, rights, and locator review. This is an organizational editorial method, not an external subject-matter credential.","targetSha256":"sha256:369c160f224159778ec08f586366d8c163e81afe67c769a0c5b9db269a01d54c","supersedesReviewId":null,"reviewerProfileVersion":2},{"scope":"domain-fidelity","verdict":"approve","reviewId":"epireview_736cbe6f7cb64d4b8dd7dea36c3783dd","rationale":"The mechanism is kept as a sparse autoencoder trained on language-model activations, and is not presented as a general dictionary-learning result or as a validated theory of superposition. This approval is limited to domain-fidelity, record urn:maha:record:mechanistic-interpretability-sae-encoder-decoder, and exact target sha256:369c160f224159778ec08f586366d8c163e81afe67c769a0c5b9db269a01d54c; it is AI-assisted internal editorial review, not external expert review, peer review, consensus, independent reproduction, scientific validation, or commercial certification.","reviewedAt":"2026-08-31T04:41:56.500Z","reviewerId":"expert_maha-internal-editorial-v2","reviewMethod":"Record-specific exact-revision review against the inspected source location, bounded claim, non-claims, rights basis, and source identity. No external reviewer participated.","reviewerKind":"internal-editorial","reviewerRole":"AI-assisted internal source, scope, boundary, rights, and locator review. This is an organizational editorial method, not an external subject-matter credential.","targetSha256":"sha256:369c160f224159778ec08f586366d8c163e81afe67c769a0c5b9db269a01d54c","supersedesReviewId":null,"reviewerProfileVersion":2},{"scope":"boundary-adequacy","verdict":"approve","reviewId":"epireview_736ab49dbc05451aa8422624397c21ba","rationale":"The record carries the source boundary that sparse features and human labels establish neither completeness, unique decomposition, nor causal faithfulness. This approval is limited to boundary-adequacy, record urn:maha:record:mechanistic-interpretability-sae-encoder-decoder, and exact target sha256:369c160f224159778ec08f586366d8c163e81afe67c769a0c5b9db269a01d54c; it is AI-assisted internal editorial review, not external expert review, peer review, consensus, independent reproduction, scientific validation, or commercial certification.","reviewedAt":"2026-08-31T04:41:56.665Z","reviewerId":"expert_maha-internal-editorial-v2","reviewMethod":"Record-specific exact-revision review against the inspected source location, bounded claim, non-claims, rights basis, and source identity. No external reviewer participated.","reviewerKind":"internal-editorial","reviewerRole":"AI-assisted internal source, scope, boundary, rights, and locator review. This is an organizational editorial method, not an external subject-matter credential.","targetSha256":"sha256:369c160f224159778ec08f586366d8c163e81afe67c769a0c5b9db269a01d54c","supersedesReviewId":null,"reviewerProfileVersion":2},{"scope":"rights-and-locator","verdict":"approve","reviewId":"epireview_b4fafaa1df9e46238dc2ad27883413a0","rationale":"arXiv:2309.08600 is cited at its method, objectives, experiments, feature-analysis and limitations sections under citation-with-paraphrase, with the preprint status of the artifact disclosed. This approval is limited to rights-and-locator, record urn:maha:record:mechanistic-interpretability-sae-encoder-decoder, and exact target sha256:369c160f224159778ec08f586366d8c163e81afe67c769a0c5b9db269a01d54c; it is AI-assisted internal editorial review, not external expert review, peer review, consensus, independent reproduction, scientific validation, or commercial certification.","reviewedAt":"2026-08-31T04:41:56.923Z","reviewerId":"expert_maha-internal-editorial-v2","reviewMethod":"Record-specific exact-revision review against the inspected source location, bounded claim, non-claims, rights basis, and source identity. No external reviewer participated.","reviewerKind":"internal-editorial","reviewerRole":"AI-assisted internal source, scope, boundary, rights, and locator review. This is an organizational editorial method, not an external subject-matter credential.","targetSha256":"sha256:369c160f224159778ec08f586366d8c163e81afe67c769a0c5b9db269a01d54c","supersedesReviewId":null,"reviewerProfileVersion":2}]}