{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:VDUVT3T63W225YQ55LTAL2VQ4D","short_pith_number":"pith:VDUVT3T6","canonical_record":{"source":{"id":"2501.19406","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-01-31T18:59:16Z","cross_cats_sorted":[],"title_canon_sha256":"827c9cd142f925ef2cafeb40e3e193826aab8eb987f3eeac21e42ed084e7abb9","abstract_canon_sha256":"9ffcdf01a24a34b13d9baf6eae2b82b1930b192ecff69d12de9dda62d391f0e7"},"schema_version":"1.0"},"canonical_sha256":"a8e959ee7eddb5aee21deae605eab0e0d10fd6c16872e3ea66b936e6ef63d010","source":{"kind":"arxiv","id":"2501.19406","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2501.19406","created_at":"2026-07-05T11:10:04Z"},{"alias_kind":"arxiv_version","alias_value":"2501.19406v2","created_at":"2026-07-05T11:10:04Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.19406","created_at":"2026-07-05T11:10:04Z"},{"alias_kind":"pith_short_12","alias_value":"VDUVT3T63W22","created_at":"2026-07-05T11:10:04Z"},{"alias_kind":"pith_short_16","alias_value":"VDUVT3T63W225YQ5","created_at":"2026-07-05T11:10:04Z"},{"alias_kind":"pith_short_8","alias_value":"VDUVT3T6","created_at":"2026-07-05T11:10:04Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:VDUVT3T63W225YQ55LTAL2VQ4D","target":"record","payload":{"canonical_record":{"source":{"id":"2501.19406","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-01-31T18:59:16Z","cross_cats_sorted":[],"title_canon_sha256":"827c9cd142f925ef2cafeb40e3e193826aab8eb987f3eeac21e42ed084e7abb9","abstract_canon_sha256":"9ffcdf01a24a34b13d9baf6eae2b82b1930b192ecff69d12de9dda62d391f0e7"},"schema_version":"1.0"},"canonical_sha256":"a8e959ee7eddb5aee21deae605eab0e0d10fd6c16872e3ea66b936e6ef63d010","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:10:04.386230Z","signature_b64":"QOkqp0hVFIuaUhTRpcFHRZ0mfuTvRZgZDIL/poxvqOvi34+lDHZyhXBfsGr6WkoXOGplJAiAnTb04kZQRz5SCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a8e959ee7eddb5aee21deae605eab0e0d10fd6c16872e3ea66b936e6ef63d010","last_reissued_at":"2026-07-05T11:10:04.385699Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:10:04.385699Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2501.19406","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:10:04Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"9SbLc3bsE+TIUg4JjYX4afJygiYlg8TzrmS5LJ45kVeYMA03h+9v+ZGfr4VkiIq+V57/u7XvQICD85UKsHuIDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T12:53:35.433819Z"},"content_sha256":"69f3613c794afa357f26503f745836bbcd453f610eed2ea97fd53d5b541dc784","schema_version":"1.0","event_id":"sha256:69f3613c794afa357f26503f745836bbcd453f610eed2ea97fd53d5b541dc784"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:VDUVT3T63W225YQ55LTAL2VQ4D","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Low-Rank Adapting Models for Sparse Autoencoders","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Joshua Engels, Matthew Chen, Max Tegmark","submitted_at":"2025-01-31T18:59:16Z","abstract_excerpt":"Sparse autoencoders (SAEs) decompose language model representations into a sparse set of linear latent vectors. Recent works have improved SAEs using language model gradients, but these techniques require many expensive backward passes during training and still cause a significant increase in cross entropy loss when SAE reconstructions are inserted into the model. In this work, we improve on these limitations by taking a fundamentally different approach: we use low-rank adaptation (LoRA) to finetune the \\textit{language model itself} around a previously trained SAE. We analyze our method acros"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.19406","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2501.19406/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:10:04Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"o+eCp8+bBGb6lci9ZnUTtg2UJwQFTkuOIefaflTihQf4+EqSJJ2b//9HOceDMtXUfKHD3FQN1gtX40abIhI9Aw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-08T12:53:35.434313Z"},"content_sha256":"040c9e67cbf8d9e72be35a0590d78cb0754fb520440f7587fd0643ddec8628ee","schema_version":"1.0","event_id":"sha256:040c9e67cbf8d9e72be35a0590d78cb0754fb520440f7587fd0643ddec8628ee"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/VDUVT3T63W225YQ55LTAL2VQ4D/bundle.json","state_url":"https://pith.science/pith/VDUVT3T63W225YQ55LTAL2VQ4D/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/VDUVT3T63W225YQ55LTAL2VQ4D/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-08T12:53:35Z","links":{"resolver":"https://pith.science/pith/VDUVT3T63W225YQ55LTAL2VQ4D","bundle":"https://pith.science/pith/VDUVT3T63W225YQ55LTAL2VQ4D/bundle.json","state":"https://pith.science/pith/VDUVT3T63W225YQ55LTAL2VQ4D/state.json","well_known_bundle":"https://pith.science/.well-known/pith/VDUVT3T63W225YQ55LTAL2VQ4D/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:VDUVT3T63W225YQ55LTAL2VQ4D","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"9ffcdf01a24a34b13d9baf6eae2b82b1930b192ecff69d12de9dda62d391f0e7","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-01-31T18:59:16Z","title_canon_sha256":"827c9cd142f925ef2cafeb40e3e193826aab8eb987f3eeac21e42ed084e7abb9"},"schema_version":"1.0","source":{"id":"2501.19406","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2501.19406","created_at":"2026-07-05T11:10:04Z"},{"alias_kind":"arxiv_version","alias_value":"2501.19406v2","created_at":"2026-07-05T11:10:04Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.19406","created_at":"2026-07-05T11:10:04Z"},{"alias_kind":"pith_short_12","alias_value":"VDUVT3T63W22","created_at":"2026-07-05T11:10:04Z"},{"alias_kind":"pith_short_16","alias_value":"VDUVT3T63W225YQ5","created_at":"2026-07-05T11:10:04Z"},{"alias_kind":"pith_short_8","alias_value":"VDUVT3T6","created_at":"2026-07-05T11:10:04Z"}],"graph_snapshots":[{"event_id":"sha256:040c9e67cbf8d9e72be35a0590d78cb0754fb520440f7587fd0643ddec8628ee","target":"graph","created_at":"2026-07-05T11:10:04Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2501.19406/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Sparse autoencoders (SAEs) decompose language model representations into a sparse set of linear latent vectors. Recent works have improved SAEs using language model gradients, but these techniques require many expensive backward passes during training and still cause a significant increase in cross entropy loss when SAE reconstructions are inserted into the model. In this work, we improve on these limitations by taking a fundamentally different approach: we use low-rank adaptation (LoRA) to finetune the \\textit{language model itself} around a previously trained SAE. We analyze our method acros","authors_text":"Joshua Engels, Matthew Chen, Max Tegmark","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-01-31T18:59:16Z","title":"Low-Rank Adapting Models for Sparse Autoencoders"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.19406","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:69f3613c794afa357f26503f745836bbcd453f610eed2ea97fd53d5b541dc784","target":"record","created_at":"2026-07-05T11:10:04Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"9ffcdf01a24a34b13d9baf6eae2b82b1930b192ecff69d12de9dda62d391f0e7","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-01-31T18:59:16Z","title_canon_sha256":"827c9cd142f925ef2cafeb40e3e193826aab8eb987f3eeac21e42ed084e7abb9"},"schema_version":"1.0","source":{"id":"2501.19406","kind":"arxiv","version":2}},"canonical_sha256":"a8e959ee7eddb5aee21deae605eab0e0d10fd6c16872e3ea66b936e6ef63d010","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a8e959ee7eddb5aee21deae605eab0e0d10fd6c16872e3ea66b936e6ef63d010","first_computed_at":"2026-07-05T11:10:04.385699Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:10:04.385699Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"QOkqp0hVFIuaUhTRpcFHRZ0mfuTvRZgZDIL/poxvqOvi34+lDHZyhXBfsGr6WkoXOGplJAiAnTb04kZQRz5SCQ==","signature_status":"signed_v1","signed_at":"2026-07-05T11:10:04.386230Z","signed_message":"canonical_sha256_bytes"},"source_id":"2501.19406","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:69f3613c794afa357f26503f745836bbcd453f610eed2ea97fd53d5b541dc784","sha256:040c9e67cbf8d9e72be35a0590d78cb0754fb520440f7587fd0643ddec8628ee"],"state_sha256":"a5392a9e86540a333b7c14454ed49531bb7507dc4210d8ddaaf0806372c5e00c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"4nqEvXSj3W/gNFKT/ucfxRSVE0okeeKOj1n/bNdRQ4qMf9hhKt2ghZHswLAYieCxwgZfy5BQ5tj+voo4UnjhCg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-08T12:53:35.437890Z","bundle_sha256":"a6eef3087309a6b58c82e735f8fedc1e1829f73acc4929bd9c81f59d7ccd4945"}}