{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:3MYT5KLZSGZHH4NKYWIVALH2L3","short_pith_number":"pith:3MYT5KLZ","canonical_record":{"source":{"id":"2410.13928","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-17T17:56:01Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"2624aae6b6c6554bb839ded4c822b2c866c91ee7602169ba2b31e2ed33681181","abstract_canon_sha256":"09976c11e711bdcac16da1422a9dcbc518a5d204462b1876e0fab8e7f6bc5eb3"},"schema_version":"1.0"},"canonical_sha256":"db313ea97991b273f1aac591502cfa5eef2159f0dbc64c213f7855da2362f275","source":{"kind":"arxiv","id":"2410.13928","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.13928","created_at":"2026-07-05T11:49:31Z"},{"alias_kind":"arxiv_version","alias_value":"2410.13928v3","created_at":"2026-07-05T11:49:31Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.13928","created_at":"2026-07-05T11:49:31Z"},{"alias_kind":"pith_short_12","alias_value":"3MYT5KLZSGZH","created_at":"2026-07-05T11:49:31Z"},{"alias_kind":"pith_short_16","alias_value":"3MYT5KLZSGZHH4NK","created_at":"2026-07-05T11:49:31Z"},{"alias_kind":"pith_short_8","alias_value":"3MYT5KLZ","created_at":"2026-07-05T11:49:31Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:3MYT5KLZSGZHH4NKYWIVALH2L3","target":"record","payload":{"canonical_record":{"source":{"id":"2410.13928","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-17T17:56:01Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"2624aae6b6c6554bb839ded4c822b2c866c91ee7602169ba2b31e2ed33681181","abstract_canon_sha256":"09976c11e711bdcac16da1422a9dcbc518a5d204462b1876e0fab8e7f6bc5eb3"},"schema_version":"1.0"},"canonical_sha256":"db313ea97991b273f1aac591502cfa5eef2159f0dbc64c213f7855da2362f275","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:49:31.868294Z","signature_b64":"ZYxfw35aJdWphr8uKhY5OD2sJq3nNU4Z1nz4TBFR0BdaHZX+L54qQNimpzJXoU8AwsnnP/4Rx/28oRukoxpuBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"db313ea97991b273f1aac591502cfa5eef2159f0dbc64c213f7855da2362f275","last_reissued_at":"2026-07-05T11:49:31.867817Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:49:31.867817Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2410.13928","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:49:31Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"gU1FO4Kv+Mdx+YD+1oIjZZqguLlYvwUGpQocW/t4xdKKjjvt7F5P1514hSdgloKIsD2ntPn7hP2Ep43E7I7CAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T16:22:58.470798Z"},"content_sha256":"254c3f1c6409202abc928ecd98f76f8783d4bcd6310721be052e266fe8e85b53","schema_version":"1.0","event_id":"sha256:254c3f1c6409202abc928ecd98f76f8783d4bcd6310721be052e266fe8e85b53"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:3MYT5KLZSGZHH4NKYWIVALH2L3","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Automatically Interpreting Millions of Features in Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Alex Mallen, Caden Juang, Gon\\c{c}alo Paulo, Nora Belrose","submitted_at":"2024-10-17T17:56:01Z","abstract_excerpt":"While the activations of neurons in deep neural networks usually do not have a simple human-understandable interpretation, sparse autoencoders (SAEs) can be used to transform these activations into a higher-dimensional latent space which may be more easily interpretable. However, these SAEs can have millions of distinct latent features, making it infeasible for humans to manually interpret each one. In this work, we build an open-source automated pipeline to generate and evaluate natural language explanations for SAE features using LLMs. We test our framework on SAEs of varying sizes, activati"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.13928","kind":"arxiv","version":3},"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/2410.13928/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":"2026-07-05T11:49:31Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6BzB31aTCYPm7iy9q5lHju2Q0yh+EauPURTEb+G4XAwJieUkKg9Kp4W9sdcNKkENiBL+5X2IKYmJT5qaz0dPDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T16:22:58.471988Z"},"content_sha256":"fd84a0db7035bf6ba38870a1cb7ab0497a628141ee8be22de2b980ba8b09f43e","schema_version":"1.0","event_id":"sha256:fd84a0db7035bf6ba38870a1cb7ab0497a628141ee8be22de2b980ba8b09f43e"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/3MYT5KLZSGZHH4NKYWIVALH2L3/bundle.json","state_url":"https://pith.science/pith/3MYT5KLZSGZHH4NKYWIVALH2L3/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/3MYT5KLZSGZHH4NKYWIVALH2L3/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":"2026-08-04T16:22:58Z","links":{"resolver":"https://pith.science/pith/3MYT5KLZSGZHH4NKYWIVALH2L3","bundle":"https://pith.science/pith/3MYT5KLZSGZHH4NKYWIVALH2L3/bundle.json","state":"https://pith.science/pith/3MYT5KLZSGZHH4NKYWIVALH2L3/state.json","well_known_bundle":"https://pith.science/.well-known/pith/3MYT5KLZSGZHH4NKYWIVALH2L3/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:3MYT5KLZSGZHH4NKYWIVALH2L3","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":"09976c11e711bdcac16da1422a9dcbc518a5d204462b1876e0fab8e7f6bc5eb3","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-17T17:56:01Z","title_canon_sha256":"2624aae6b6c6554bb839ded4c822b2c866c91ee7602169ba2b31e2ed33681181"},"schema_version":"1.0","source":{"id":"2410.13928","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2410.13928","created_at":"2026-07-05T11:49:31Z"},{"alias_kind":"arxiv_version","alias_value":"2410.13928v3","created_at":"2026-07-05T11:49:31Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.13928","created_at":"2026-07-05T11:49:31Z"},{"alias_kind":"pith_short_12","alias_value":"3MYT5KLZSGZH","created_at":"2026-07-05T11:49:31Z"},{"alias_kind":"pith_short_16","alias_value":"3MYT5KLZSGZHH4NK","created_at":"2026-07-05T11:49:31Z"},{"alias_kind":"pith_short_8","alias_value":"3MYT5KLZ","created_at":"2026-07-05T11:49:31Z"}],"graph_snapshots":[{"event_id":"sha256:fd84a0db7035bf6ba38870a1cb7ab0497a628141ee8be22de2b980ba8b09f43e","target":"graph","created_at":"2026-07-05T11:49:31Z","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/2410.13928/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"While the activations of neurons in deep neural networks usually do not have a simple human-understandable interpretation, sparse autoencoders (SAEs) can be used to transform these activations into a higher-dimensional latent space which may be more easily interpretable. However, these SAEs can have millions of distinct latent features, making it infeasible for humans to manually interpret each one. In this work, we build an open-source automated pipeline to generate and evaluate natural language explanations for SAE features using LLMs. We test our framework on SAEs of varying sizes, activati","authors_text":"Alex Mallen, Caden Juang, Gon\\c{c}alo Paulo, Nora Belrose","cross_cats":["cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-17T17:56:01Z","title":"Automatically Interpreting Millions of Features in Large Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.13928","kind":"arxiv","version":3},"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":"sha256:254c3f1c6409202abc928ecd98f76f8783d4bcd6310721be052e266fe8e85b53","target":"record","created_at":"2026-07-05T11:49:31Z","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":"09976c11e711bdcac16da1422a9dcbc518a5d204462b1876e0fab8e7f6bc5eb3","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-17T17:56:01Z","title_canon_sha256":"2624aae6b6c6554bb839ded4c822b2c866c91ee7602169ba2b31e2ed33681181"},"schema_version":"1.0","source":{"id":"2410.13928","kind":"arxiv","version":3}},"canonical_sha256":"db313ea97991b273f1aac591502cfa5eef2159f0dbc64c213f7855da2362f275","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"db313ea97991b273f1aac591502cfa5eef2159f0dbc64c213f7855da2362f275","first_computed_at":"2026-07-05T11:49:31.867817Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:49:31.867817Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ZYxfw35aJdWphr8uKhY5OD2sJq3nNU4Z1nz4TBFR0BdaHZX+L54qQNimpzJXoU8AwsnnP/4Rx/28oRukoxpuBg==","signature_status":"signed_v1","signed_at":"2026-07-05T11:49:31.868294Z","signed_message":"canonical_sha256_bytes"},"source_id":"2410.13928","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:254c3f1c6409202abc928ecd98f76f8783d4bcd6310721be052e266fe8e85b53","sha256:fd84a0db7035bf6ba38870a1cb7ab0497a628141ee8be22de2b980ba8b09f43e"],"state_sha256":"14a420be5bb47d524a3905b9375ed12d9e6f6c944677d7882226ea4e48648a8a"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"CIYkuO83kXrR9guIs+Pn8r1QTOpp/VNCrwORbg2g/qEg/CZ5rfmoICF4CiB6876CkE/cP0B4VD/HDtv7JcM5Aw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T16:22:58.512246Z","bundle_sha256":"5825c7ae55b8255d3bbdb63d2375239774bd4c4ed9949e7178c52adfd631769f"}}