{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:5IEZ3NVZL66JJ2TMJZN5KSLNYA","short_pith_number":"pith:5IEZ3NVZ","canonical_record":{"source":{"id":"2505.20254","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-26T17:31:36Z","cross_cats_sorted":["cs.AI","cs.CL","stat.ML"],"title_canon_sha256":"cd36ea922881121b32c02889a1df6e8f5230fd5d8b17f2ed2e82cd47f24fcc15","abstract_canon_sha256":"780becec68f53c1d2a56269217b74cac22ee975d74fd2677631c726ae975aca0"},"schema_version":"1.0"},"canonical_sha256":"ea099db6b95fbc94ea6c4e5bd5496dc032275bf6a3d15bc54e862ba340346331","source":{"kind":"arxiv","id":"2505.20254","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.20254","created_at":"2026-07-05T11:09:55Z"},{"alias_kind":"arxiv_version","alias_value":"2505.20254v1","created_at":"2026-07-05T11:09:55Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.20254","created_at":"2026-07-05T11:09:55Z"},{"alias_kind":"pith_short_12","alias_value":"5IEZ3NVZL66J","created_at":"2026-07-05T11:09:55Z"},{"alias_kind":"pith_short_16","alias_value":"5IEZ3NVZL66JJ2TM","created_at":"2026-07-05T11:09:55Z"},{"alias_kind":"pith_short_8","alias_value":"5IEZ3NVZ","created_at":"2026-07-05T11:09:55Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:5IEZ3NVZL66JJ2TMJZN5KSLNYA","target":"record","payload":{"canonical_record":{"source":{"id":"2505.20254","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-26T17:31:36Z","cross_cats_sorted":["cs.AI","cs.CL","stat.ML"],"title_canon_sha256":"cd36ea922881121b32c02889a1df6e8f5230fd5d8b17f2ed2e82cd47f24fcc15","abstract_canon_sha256":"780becec68f53c1d2a56269217b74cac22ee975d74fd2677631c726ae975aca0"},"schema_version":"1.0"},"canonical_sha256":"ea099db6b95fbc94ea6c4e5bd5496dc032275bf6a3d15bc54e862ba340346331","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:09:55.192497Z","signature_b64":"saJSWXAHNMrOwkSWHZA2ezPqPdyWvbJJEdxC3j0Zbk0hsn7LSCm4KSjw0O1UMN3RFQnF1DZ5p5JPJ7j2MrPQAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ea099db6b95fbc94ea6c4e5bd5496dc032275bf6a3d15bc54e862ba340346331","last_reissued_at":"2026-07-05T11:09:55.192005Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:09:55.192005Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2505.20254","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":"2026-07-05T11:09:55Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"YbZzkVYbDyfYY1klj2c8hnHtgXBAMrErgyhQVYJGaqZUF35qEtPRp4Q0lKCmGWz6Pc1r/9ySR0DoVC1ESVVFDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-20T12:36:14.776782Z"},"content_sha256":"0008fcd4acc7a7aa15581078bd48f90da2f1b0056fc67f1500234214ea46203f","schema_version":"1.0","event_id":"sha256:0008fcd4acc7a7aa15581078bd48f90da2f1b0056fc67f1500234214ea46203f"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:5IEZ3NVZL66JJ2TMJZN5KSLNYA","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Aashiq Muhamed, Kun Zhang, Lingjing Kong, Mona T. Diab, Virginia Smith, Xiangchen Song, Yujia Zheng, Zeyu Tang","submitted_at":"2025-05-26T17:31:36Z","abstract_excerpt":"Sparse Autoencoders (SAEs) are a prominent tool in mechanistic interpretability (MI) for decomposing neural network activations into interpretable features. However, the aspiration to identify a canonical set of features is challenged by the observed inconsistency of learned SAE features across different training runs, undermining the reliability and efficiency of MI research. This position paper argues that mechanistic interpretability should prioritize feature consistency in SAEs -- the reliable convergence to equivalent feature sets across independent runs. We propose using the Pairwise Dic"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.20254","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/2505.20254/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":"2026-07-05T11:09:55Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1BdBsxNc8/PsGLK9WU1yG92DrvFbWrfh0hiqNIzqviow1ilZVM8vnk0/WCFveovDDni+3hd0OAaFFKd4g5MTAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-20T12:36:14.777313Z"},"content_sha256":"91e8dd320273f3a1095ebc5f046ff8d16d811eb3ede9806e42d6f0fda4d743ff","schema_version":"1.0","event_id":"sha256:91e8dd320273f3a1095ebc5f046ff8d16d811eb3ede9806e42d6f0fda4d743ff"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/5IEZ3NVZL66JJ2TMJZN5KSLNYA/bundle.json","state_url":"https://pith.science/pith/5IEZ3NVZL66JJ2TMJZN5KSLNYA/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/5IEZ3NVZL66JJ2TMJZN5KSLNYA/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":"2026-08-20T12:36:14Z","links":{"resolver":"https://pith.science/pith/5IEZ3NVZL66JJ2TMJZN5KSLNYA","bundle":"https://pith.science/pith/5IEZ3NVZL66JJ2TMJZN5KSLNYA/bundle.json","state":"https://pith.science/pith/5IEZ3NVZL66JJ2TMJZN5KSLNYA/state.json","well_known_bundle":"https://pith.science/.well-known/pith/5IEZ3NVZL66JJ2TMJZN5KSLNYA/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:5IEZ3NVZL66JJ2TMJZN5KSLNYA","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":"780becec68f53c1d2a56269217b74cac22ee975d74fd2677631c726ae975aca0","cross_cats_sorted":["cs.AI","cs.CL","stat.ML"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-26T17:31:36Z","title_canon_sha256":"cd36ea922881121b32c02889a1df6e8f5230fd5d8b17f2ed2e82cd47f24fcc15"},"schema_version":"1.0","source":{"id":"2505.20254","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.20254","created_at":"2026-07-05T11:09:55Z"},{"alias_kind":"arxiv_version","alias_value":"2505.20254v1","created_at":"2026-07-05T11:09:55Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.20254","created_at":"2026-07-05T11:09:55Z"},{"alias_kind":"pith_short_12","alias_value":"5IEZ3NVZL66J","created_at":"2026-07-05T11:09:55Z"},{"alias_kind":"pith_short_16","alias_value":"5IEZ3NVZL66JJ2TM","created_at":"2026-07-05T11:09:55Z"},{"alias_kind":"pith_short_8","alias_value":"5IEZ3NVZ","created_at":"2026-07-05T11:09:55Z"}],"graph_snapshots":[{"event_id":"sha256:91e8dd320273f3a1095ebc5f046ff8d16d811eb3ede9806e42d6f0fda4d743ff","target":"graph","created_at":"2026-07-05T11:09:55Z","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/2505.20254/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Sparse Autoencoders (SAEs) are a prominent tool in mechanistic interpretability (MI) for decomposing neural network activations into interpretable features. However, the aspiration to identify a canonical set of features is challenged by the observed inconsistency of learned SAE features across different training runs, undermining the reliability and efficiency of MI research. This position paper argues that mechanistic interpretability should prioritize feature consistency in SAEs -- the reliable convergence to equivalent feature sets across independent runs. We propose using the Pairwise Dic","authors_text":"Aashiq Muhamed, Kun Zhang, Lingjing Kong, Mona T. Diab, Virginia Smith, Xiangchen Song, Yujia Zheng, Zeyu Tang","cross_cats":["cs.AI","cs.CL","stat.ML"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-26T17:31:36Z","title":"Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.20254","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":"sha256:0008fcd4acc7a7aa15581078bd48f90da2f1b0056fc67f1500234214ea46203f","target":"record","created_at":"2026-07-05T11:09:55Z","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":"780becec68f53c1d2a56269217b74cac22ee975d74fd2677631c726ae975aca0","cross_cats_sorted":["cs.AI","cs.CL","stat.ML"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-26T17:31:36Z","title_canon_sha256":"cd36ea922881121b32c02889a1df6e8f5230fd5d8b17f2ed2e82cd47f24fcc15"},"schema_version":"1.0","source":{"id":"2505.20254","kind":"arxiv","version":1}},"canonical_sha256":"ea099db6b95fbc94ea6c4e5bd5496dc032275bf6a3d15bc54e862ba340346331","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"ea099db6b95fbc94ea6c4e5bd5496dc032275bf6a3d15bc54e862ba340346331","first_computed_at":"2026-07-05T11:09:55.192005Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:09:55.192005Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"saJSWXAHNMrOwkSWHZA2ezPqPdyWvbJJEdxC3j0Zbk0hsn7LSCm4KSjw0O1UMN3RFQnF1DZ5p5JPJ7j2MrPQAg==","signature_status":"signed_v1","signed_at":"2026-07-05T11:09:55.192497Z","signed_message":"canonical_sha256_bytes"},"source_id":"2505.20254","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:0008fcd4acc7a7aa15581078bd48f90da2f1b0056fc67f1500234214ea46203f","sha256:91e8dd320273f3a1095ebc5f046ff8d16d811eb3ede9806e42d6f0fda4d743ff"],"state_sha256":"6aa6767cf74d528cfb455d4bad2bccbe632c71425affed5681f5223906523095"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"wK1kOooCgH/Ax2gq1xxOKb+jmTQEuzg7ClhXZTCdDTShJlrRzLkI5J6S6PmzQwlLt4t8/JBYWVWUZY+6iPF6DA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-20T12:36:14.782390Z","bundle_sha256":"e0d9bd8bb1835b158ec14d8d6218116cac3309982144638e1156d0ab99c06ec0"}}