{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:6INO655OCX65GAVNQTOXMX22MT","short_pith_number":"pith:6INO655O","canonical_record":{"source":{"id":"2411.18895","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-11-28T03:58:48Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"658586de7bf322a9dacf226dc73d8801d749b72572c851092d5134dc4ec5120c","abstract_canon_sha256":"4326b5137cd0de75e27dd7b785893b6ace4d143805025146feb7aab552373e7a"},"schema_version":"1.0"},"canonical_sha256":"f21aef77ae15fdd302ad84dd765f5a64c818e43398ee629d50b9928c6ae8e75a","source":{"kind":"arxiv","id":"2411.18895","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2411.18895","created_at":"2026-07-05T09:41:53Z"},{"alias_kind":"arxiv_version","alias_value":"2411.18895v1","created_at":"2026-07-05T09:41:53Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.18895","created_at":"2026-07-05T09:41:53Z"},{"alias_kind":"pith_short_12","alias_value":"6INO655OCX65","created_at":"2026-07-05T09:41:53Z"},{"alias_kind":"pith_short_16","alias_value":"6INO655OCX65GAVN","created_at":"2026-07-05T09:41:53Z"},{"alias_kind":"pith_short_8","alias_value":"6INO655O","created_at":"2026-07-05T09:41:53Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:6INO655OCX65GAVNQTOXMX22MT","target":"record","payload":{"canonical_record":{"source":{"id":"2411.18895","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-11-28T03:58:48Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"658586de7bf322a9dacf226dc73d8801d749b72572c851092d5134dc4ec5120c","abstract_canon_sha256":"4326b5137cd0de75e27dd7b785893b6ace4d143805025146feb7aab552373e7a"},"schema_version":"1.0"},"canonical_sha256":"f21aef77ae15fdd302ad84dd765f5a64c818e43398ee629d50b9928c6ae8e75a","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:41:53.364158Z","signature_b64":"15B8g3aYKeQje7Y5UDRqO62FBTRiFD5nu7xqOPiAbMszqPpxXzZUzhhFL87xKBT5SXYviaZbTBcHANEaHBNUBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f21aef77ae15fdd302ad84dd765f5a64c818e43398ee629d50b9928c6ae8e75a","last_reissued_at":"2026-07-05T09:41:53.363675Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:41:53.363675Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2411.18895","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":"2026-07-05T09:41:53Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"hVqGK6/h1AGbllEpQEVwHXCowFxzhcjXUFLcW15/BR8CR2cT2OsFRMdnDLa+equXtnJiCXP4BclXmf0E3PP7DA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T17:05:48.469880Z"},"content_sha256":"d96bf877a35da8d2e3e781d77c677b607e74c90db5117c793d281079d058c356","schema_version":"1.0","event_id":"sha256:d96bf877a35da8d2e3e781d77c677b607e74c90db5117c793d281079d058c356"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:6INO655OCX65GAVNQTOXMX22MT","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Evaluating Sparse Autoencoders on Targeted Concept Erasure Tasks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Adam Karvonen, Can Rager, Neel Nanda, Samuel Marks","submitted_at":"2024-11-28T03:58:48Z","abstract_excerpt":"Sparse Autoencoders (SAEs) are an interpretability technique aimed at decomposing neural network activations into interpretable units. However, a major bottleneck for SAE development has been the lack of high-quality performance metrics, with prior work largely relying on unsupervised proxies. In this work, we introduce a family of evaluations based on SHIFT, a downstream task from Marks et al. (Sparse Feature Circuits, 2024) in which spurious cues are removed from a classifier by ablating SAE features judged to be task-irrelevant by a human annotator. We adapt SHIFT into an automated metric o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.18895","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/2411.18895/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":"2026-07-05T09:41:53Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"VGsuqpV1uqUMEsHPv+b2tcAyHmN+Ynjz3cBlzB0Ki+JBVAhFWuC7dfw4MMU0SKBV6Gs/VRoB/hKk2N5Njeb4Dw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T17:05:48.470964Z"},"content_sha256":"03692cab9616b440fa2ceac6d090337b1fc2bc96848289266d63da80a960aa21","schema_version":"1.0","event_id":"sha256:03692cab9616b440fa2ceac6d090337b1fc2bc96848289266d63da80a960aa21"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/6INO655OCX65GAVNQTOXMX22MT/bundle.json","state_url":"https://pith.science/pith/6INO655OCX65GAVNQTOXMX22MT/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/6INO655OCX65GAVNQTOXMX22MT/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":"2026-08-10T17:05:48Z","links":{"resolver":"https://pith.science/pith/6INO655OCX65GAVNQTOXMX22MT","bundle":"https://pith.science/pith/6INO655OCX65GAVNQTOXMX22MT/bundle.json","state":"https://pith.science/pith/6INO655OCX65GAVNQTOXMX22MT/state.json","well_known_bundle":"https://pith.science/.well-known/pith/6INO655OCX65GAVNQTOXMX22MT/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:6INO655OCX65GAVNQTOXMX22MT","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":"4326b5137cd0de75e27dd7b785893b6ace4d143805025146feb7aab552373e7a","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-11-28T03:58:48Z","title_canon_sha256":"658586de7bf322a9dacf226dc73d8801d749b72572c851092d5134dc4ec5120c"},"schema_version":"1.0","source":{"id":"2411.18895","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2411.18895","created_at":"2026-07-05T09:41:53Z"},{"alias_kind":"arxiv_version","alias_value":"2411.18895v1","created_at":"2026-07-05T09:41:53Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.18895","created_at":"2026-07-05T09:41:53Z"},{"alias_kind":"pith_short_12","alias_value":"6INO655OCX65","created_at":"2026-07-05T09:41:53Z"},{"alias_kind":"pith_short_16","alias_value":"6INO655OCX65GAVN","created_at":"2026-07-05T09:41:53Z"},{"alias_kind":"pith_short_8","alias_value":"6INO655O","created_at":"2026-07-05T09:41:53Z"}],"graph_snapshots":[{"event_id":"sha256:03692cab9616b440fa2ceac6d090337b1fc2bc96848289266d63da80a960aa21","target":"graph","created_at":"2026-07-05T09:41:53Z","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/2411.18895/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Sparse Autoencoders (SAEs) are an interpretability technique aimed at decomposing neural network activations into interpretable units. However, a major bottleneck for SAE development has been the lack of high-quality performance metrics, with prior work largely relying on unsupervised proxies. In this work, we introduce a family of evaluations based on SHIFT, a downstream task from Marks et al. (Sparse Feature Circuits, 2024) in which spurious cues are removed from a classifier by ablating SAE features judged to be task-irrelevant by a human annotator. We adapt SHIFT into an automated metric o","authors_text":"Adam Karvonen, Can Rager, Neel Nanda, Samuel Marks","cross_cats":["cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-11-28T03:58:48Z","title":"Evaluating Sparse Autoencoders on Targeted Concept Erasure Tasks"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.18895","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:d96bf877a35da8d2e3e781d77c677b607e74c90db5117c793d281079d058c356","target":"record","created_at":"2026-07-05T09:41:53Z","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":"4326b5137cd0de75e27dd7b785893b6ace4d143805025146feb7aab552373e7a","cross_cats_sorted":["cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-11-28T03:58:48Z","title_canon_sha256":"658586de7bf322a9dacf226dc73d8801d749b72572c851092d5134dc4ec5120c"},"schema_version":"1.0","source":{"id":"2411.18895","kind":"arxiv","version":1}},"canonical_sha256":"f21aef77ae15fdd302ad84dd765f5a64c818e43398ee629d50b9928c6ae8e75a","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"f21aef77ae15fdd302ad84dd765f5a64c818e43398ee629d50b9928c6ae8e75a","first_computed_at":"2026-07-05T09:41:53.363675Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T09:41:53.363675Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"15B8g3aYKeQje7Y5UDRqO62FBTRiFD5nu7xqOPiAbMszqPpxXzZUzhhFL87xKBT5SXYviaZbTBcHANEaHBNUBA==","signature_status":"signed_v1","signed_at":"2026-07-05T09:41:53.364158Z","signed_message":"canonical_sha256_bytes"},"source_id":"2411.18895","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:d96bf877a35da8d2e3e781d77c677b607e74c90db5117c793d281079d058c356","sha256:03692cab9616b440fa2ceac6d090337b1fc2bc96848289266d63da80a960aa21"],"state_sha256":"0bcb2231a5357496a29e8c9c1e6642dcb77d198442a8136595d5e409923c9eee"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"kwb3Nwm5GVGV9muqYHbGaLMis3JDETF64HNe/6zzdxYdUYcbKqCrvqY3n97/x2ZUBmD5Cw0eJFC98xuaDQXZAg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-10T17:05:48.477938Z","bundle_sha256":"2376bc2c415ec6bba4752e34152d1ade7372681a9214b396dd9c2cbd93bc5fb5"}}