{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:JD4C457K56CHCSUJSZRKKFNPYA","short_pith_number":"pith:JD4C457K","canonical_record":{"source":{"id":"2503.17547","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-03-21T21:43:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"6839c8eb1e25bf9730a4c71744cfd6126f1403107ced73d322f43f71f41a9c7f","abstract_canon_sha256":"190ef8f1c3acd4408e4113940647c808cca69168850d67ecccf74b157540fd3d"},"schema_version":"1.0"},"canonical_sha256":"48f82e77eaef84714a899662a515afc0325f928bb1a7b326f00aa269e9ca18b2","source":{"kind":"arxiv","id":"2503.17547","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2503.17547","created_at":"2026-07-05T10:37:11Z"},{"alias_kind":"arxiv_version","alias_value":"2503.17547v1","created_at":"2026-07-05T10:37:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.17547","created_at":"2026-07-05T10:37:11Z"},{"alias_kind":"pith_short_12","alias_value":"JD4C457K56CH","created_at":"2026-07-05T10:37:11Z"},{"alias_kind":"pith_short_16","alias_value":"JD4C457K56CHCSUJ","created_at":"2026-07-05T10:37:11Z"},{"alias_kind":"pith_short_8","alias_value":"JD4C457K","created_at":"2026-07-05T10:37:11Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:JD4C457K56CHCSUJSZRKKFNPYA","target":"record","payload":{"canonical_record":{"source":{"id":"2503.17547","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-03-21T21:43:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"6839c8eb1e25bf9730a4c71744cfd6126f1403107ced73d322f43f71f41a9c7f","abstract_canon_sha256":"190ef8f1c3acd4408e4113940647c808cca69168850d67ecccf74b157540fd3d"},"schema_version":"1.0"},"canonical_sha256":"48f82e77eaef84714a899662a515afc0325f928bb1a7b326f00aa269e9ca18b2","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:37:11.750052Z","signature_b64":"ZPmQKujFptlbPFoY876PPGktqgA1sfoduUcuYAAXD/Xa7zBBDEoXoFDQTUn7yb1Rvnrs4Xh2UcvAWxprgGXPDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"48f82e77eaef84714a899662a515afc0325f928bb1a7b326f00aa269e9ca18b2","last_reissued_at":"2026-07-05T10:37:11.749559Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:37:11.749559Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2503.17547","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:37:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"prP83nql9CERWXZqnPp+BT22WDoAE/jhdNDKxmEtpZMe4YR3DHS1QPexNB2UdJzD4fgFhH9MQnBj+upTEmuKCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-07T06:33:11.614593Z"},"content_sha256":"ac7baefe2fa5091c7487d40c53c3843bf157de17fa46eb40aa22498d477b6389","schema_version":"1.0","event_id":"sha256:ac7baefe2fa5091c7487d40c53c3843bf157de17fa46eb40aa22498d477b6389"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:JD4C457K56CHCSUJSZRKKFNPYA","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Learning Multi-Level Features with Matryoshka Sparse Autoencoders","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Adam Karvonen, Bart Bussmann, Neel Nanda, Noa Nabeshima","submitted_at":"2025-03-21T21:43:28Z","abstract_excerpt":"Sparse autoencoders (SAEs) have emerged as a powerful tool for interpreting neural networks by extracting the concepts represented in their activations. However, choosing the size of the SAE dictionary (i.e. number of learned concepts) creates a tension: as dictionary size increases to capture more relevant concepts, sparsity incentivizes features to be split or absorbed into more specific features, leaving high-level features missing or warped. We introduce Matryoshka SAEs, a novel variant that addresses these issues by simultaneously training multiple nested dictionaries of increasing size, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.17547","kind":"arxiv","version":1},"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/2503.17547/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":"2026-07-05T10:37:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"wrbzxPul9eScsjMtvkh0zcI6rNsCeG/D6965G+2nXpiinkHrPdJ6PgwwpYfzL/W7KpLngYWCApyRX6vBUKVoBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-07T06:33:11.615537Z"},"content_sha256":"d6566b2f52a48e593348a97876d50e018a0b1e507efd275d5b8115c3da29036b","schema_version":"1.0","event_id":"sha256:d6566b2f52a48e593348a97876d50e018a0b1e507efd275d5b8115c3da29036b"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/JD4C457K56CHCSUJSZRKKFNPYA/bundle.json","state_url":"https://pith.science/pith/JD4C457K56CHCSUJSZRKKFNPYA/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/JD4C457K56CHCSUJSZRKKFNPYA/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":"2026-08-07T06:33:11Z","links":{"resolver":"https://pith.science/pith/JD4C457K56CHCSUJSZRKKFNPYA","bundle":"https://pith.science/pith/JD4C457K56CHCSUJSZRKKFNPYA/bundle.json","state":"https://pith.science/pith/JD4C457K56CHCSUJSZRKKFNPYA/state.json","well_known_bundle":"https://pith.science/.well-known/pith/JD4C457K56CHCSUJSZRKKFNPYA/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:JD4C457K56CHCSUJSZRKKFNPYA","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":"190ef8f1c3acd4408e4113940647c808cca69168850d67ecccf74b157540fd3d","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-03-21T21:43:28Z","title_canon_sha256":"6839c8eb1e25bf9730a4c71744cfd6126f1403107ced73d322f43f71f41a9c7f"},"schema_version":"1.0","source":{"id":"2503.17547","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2503.17547","created_at":"2026-07-05T10:37:11Z"},{"alias_kind":"arxiv_version","alias_value":"2503.17547v1","created_at":"2026-07-05T10:37:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.17547","created_at":"2026-07-05T10:37:11Z"},{"alias_kind":"pith_short_12","alias_value":"JD4C457K56CH","created_at":"2026-07-05T10:37:11Z"},{"alias_kind":"pith_short_16","alias_value":"JD4C457K56CHCSUJ","created_at":"2026-07-05T10:37:11Z"},{"alias_kind":"pith_short_8","alias_value":"JD4C457K","created_at":"2026-07-05T10:37:11Z"}],"graph_snapshots":[{"event_id":"sha256:d6566b2f52a48e593348a97876d50e018a0b1e507efd275d5b8115c3da29036b","target":"graph","created_at":"2026-07-05T10:37:11Z","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/2503.17547/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Sparse autoencoders (SAEs) have emerged as a powerful tool for interpreting neural networks by extracting the concepts represented in their activations. However, choosing the size of the SAE dictionary (i.e. number of learned concepts) creates a tension: as dictionary size increases to capture more relevant concepts, sparsity incentivizes features to be split or absorbed into more specific features, leaving high-level features missing or warped. We introduce Matryoshka SAEs, a novel variant that addresses these issues by simultaneously training multiple nested dictionaries of increasing size, ","authors_text":"Adam Karvonen, Bart Bussmann, Neel Nanda, Noa Nabeshima","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-03-21T21:43:28Z","title":"Learning Multi-Level Features with Matryoshka Sparse Autoencoders"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.17547","kind":"arxiv","version":1},"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:ac7baefe2fa5091c7487d40c53c3843bf157de17fa46eb40aa22498d477b6389","target":"record","created_at":"2026-07-05T10:37:11Z","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":"190ef8f1c3acd4408e4113940647c808cca69168850d67ecccf74b157540fd3d","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-03-21T21:43:28Z","title_canon_sha256":"6839c8eb1e25bf9730a4c71744cfd6126f1403107ced73d322f43f71f41a9c7f"},"schema_version":"1.0","source":{"id":"2503.17547","kind":"arxiv","version":1}},"canonical_sha256":"48f82e77eaef84714a899662a515afc0325f928bb1a7b326f00aa269e9ca18b2","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"48f82e77eaef84714a899662a515afc0325f928bb1a7b326f00aa269e9ca18b2","first_computed_at":"2026-07-05T10:37:11.749559Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:37:11.749559Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ZPmQKujFptlbPFoY876PPGktqgA1sfoduUcuYAAXD/Xa7zBBDEoXoFDQTUn7yb1Rvnrs4Xh2UcvAWxprgGXPDg==","signature_status":"signed_v1","signed_at":"2026-07-05T10:37:11.750052Z","signed_message":"canonical_sha256_bytes"},"source_id":"2503.17547","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:ac7baefe2fa5091c7487d40c53c3843bf157de17fa46eb40aa22498d477b6389","sha256:d6566b2f52a48e593348a97876d50e018a0b1e507efd275d5b8115c3da29036b"],"state_sha256":"ddf693d0197cfbc86ae7d2fcfc87fe8d7ca76e3be74225f8cd5a721cb01deb5c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6RSbOX/vGz7Nhi70XQmzZ+RBtVWfXXazpjt+9R2gyqNXR063MhZKtXbt/KVCzOhuF1x8Xr/DwZ06/EJOi6SPCQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-07T06:33:11.620196Z","bundle_sha256":"4c8053d582df9d8b55f41b90b2c32474f0c525993f3222d6fd16fdaa07963145"}}