{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:GTZ44TTJLW3M2YULGA3MBFXOOM","short_pith_number":"pith:GTZ44TTJ","canonical_record":{"source":{"id":"2605.18229","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-18T11:20:57Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f3033cd9c6ce8b124b5a5289c0059d9d7579f3995274d3a9682ecbe0b24ced23","abstract_canon_sha256":"82bd1ecc62b69cf4b87140bdf005e06b458549039bf2966b9e4953efc9b33709"},"schema_version":"1.0"},"canonical_sha256":"34f3ce4e695db6cd628b3036c096ee730f9654436054cf37345194c0d9e8f057","source":{"kind":"arxiv","id":"2605.18229","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.18229","created_at":"2026-05-20T00:05:51Z"},{"alias_kind":"arxiv_version","alias_value":"2605.18229v1","created_at":"2026-05-20T00:05:51Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.18229","created_at":"2026-05-20T00:05:51Z"},{"alias_kind":"pith_short_12","alias_value":"GTZ44TTJLW3M","created_at":"2026-05-20T00:05:51Z"},{"alias_kind":"pith_short_16","alias_value":"GTZ44TTJLW3M2YUL","created_at":"2026-05-20T00:05:51Z"},{"alias_kind":"pith_short_8","alias_value":"GTZ44TTJ","created_at":"2026-05-20T00:05:51Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:GTZ44TTJLW3M2YULGA3MBFXOOM","target":"record","payload":{"canonical_record":{"source":{"id":"2605.18229","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-18T11:20:57Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f3033cd9c6ce8b124b5a5289c0059d9d7579f3995274d3a9682ecbe0b24ced23","abstract_canon_sha256":"82bd1ecc62b69cf4b87140bdf005e06b458549039bf2966b9e4953efc9b33709"},"schema_version":"1.0"},"canonical_sha256":"34f3ce4e695db6cd628b3036c096ee730f9654436054cf37345194c0d9e8f057","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:05:51.460441Z","signature_b64":"RkY4RnSTNdkuU53z72sIODJX7laiugsCC5u+KGs8LTSJ1taNEtA3a7tjPwHlSy5WWii7sNjJZ36RLqWUnoN3Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"34f3ce4e695db6cd628b3036c096ee730f9654436054cf37345194c0d9e8f057","last_reissued_at":"2026-05-20T00:05:51.460020Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:05:51.460020Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.18229","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":"2026-05-20T00:05:51Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"IsZXb6XqXo/+VSfIf/hH17nk32E2IYHWycegbxW99oX0mij1BegVyWbqSpAt8AcGPsJsypn7P9T7XkI0fUnNCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-25T18:03:19.659798Z"},"content_sha256":"a7c694dc96f13546b2ed01c4ba8af4dc4da0464ba17570cde9556a040dfe9446","schema_version":"1.0","event_id":"sha256:a7c694dc96f13546b2ed01c4ba8af4dc4da0464ba17570cde9556a040dfe9446"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:GTZ44TTJLW3M2YULGA3MBFXOOM","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Are Sparse Autoencoder Benchmarks Reliable?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"David Chanin","submitted_at":"2026-05-18T11:20:57Z","abstract_excerpt":"Sparse autoencoders (SAEs) are a core interpretability tool for large language models, and progress on SAE architectures depends on benchmarks that reliably distinguish better SAEs from worse ones. We audit the SAE quality metrics in SAEBench, the de-facto standard SAE evaluation suite, through three complementary lenses: reseed noise on a fixed SAE, ground-truth correlation on synthetic SAEs, and discriminability across training trajectories. We find that two of these metrics, Targeted Probe Perturbation (TPP) and Spurious Correlation Removal (SCR), fail multiple lenses at their canonical set"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.18229","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/2605.18229/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"claim_evidence","ran_at":"2026-05-19T23:41:58.937145Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T23:33:35.292780Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"e683024d011065288389d1b9b921aa34705a3a61d151cc39ce5def54529a925d"},"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":"2026-05-20T00:05:51Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"qzcdmVv6xXo5iD9XSBTuOzlpzgxaEfGFGrTB7yqHBgx7O80Yqkos4c9SMIIhw+4cbosezoe0qyJy3YwDl0o3Bw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-25T18:03:19.660574Z"},"content_sha256":"fd2758dca50324ee674a61e2bb0cc80ee94a5380944de89b3b28fd9853e94a20","schema_version":"1.0","event_id":"sha256:fd2758dca50324ee674a61e2bb0cc80ee94a5380944de89b3b28fd9853e94a20"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/GTZ44TTJLW3M2YULGA3MBFXOOM/bundle.json","state_url":"https://pith.science/pith/GTZ44TTJLW3M2YULGA3MBFXOOM/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/GTZ44TTJLW3M2YULGA3MBFXOOM/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":"2026-05-25T18:03:19Z","links":{"resolver":"https://pith.science/pith/GTZ44TTJLW3M2YULGA3MBFXOOM","bundle":"https://pith.science/pith/GTZ44TTJLW3M2YULGA3MBFXOOM/bundle.json","state":"https://pith.science/pith/GTZ44TTJLW3M2YULGA3MBFXOOM/state.json","well_known_bundle":"https://pith.science/.well-known/pith/GTZ44TTJLW3M2YULGA3MBFXOOM/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:GTZ44TTJLW3M2YULGA3MBFXOOM","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":"82bd1ecc62b69cf4b87140bdf005e06b458549039bf2966b9e4953efc9b33709","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-18T11:20:57Z","title_canon_sha256":"f3033cd9c6ce8b124b5a5289c0059d9d7579f3995274d3a9682ecbe0b24ced23"},"schema_version":"1.0","source":{"id":"2605.18229","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.18229","created_at":"2026-05-20T00:05:51Z"},{"alias_kind":"arxiv_version","alias_value":"2605.18229v1","created_at":"2026-05-20T00:05:51Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.18229","created_at":"2026-05-20T00:05:51Z"},{"alias_kind":"pith_short_12","alias_value":"GTZ44TTJLW3M","created_at":"2026-05-20T00:05:51Z"},{"alias_kind":"pith_short_16","alias_value":"GTZ44TTJLW3M2YUL","created_at":"2026-05-20T00:05:51Z"},{"alias_kind":"pith_short_8","alias_value":"GTZ44TTJ","created_at":"2026-05-20T00:05:51Z"}],"graph_snapshots":[{"event_id":"sha256:fd2758dca50324ee674a61e2bb0cc80ee94a5380944de89b3b28fd9853e94a20","target":"graph","created_at":"2026-05-20T00:05:51Z","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":[{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T23:41:58.937145Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T23:33:35.292780Z","status":"skipped","version":"1.0.0"}],"endpoint":"/pith/2605.18229/integrity.json","findings":[],"snapshot_sha256":"e683024d011065288389d1b9b921aa34705a3a61d151cc39ce5def54529a925d","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Sparse autoencoders (SAEs) are a core interpretability tool for large language models, and progress on SAE architectures depends on benchmarks that reliably distinguish better SAEs from worse ones. We audit the SAE quality metrics in SAEBench, the de-facto standard SAE evaluation suite, through three complementary lenses: reseed noise on a fixed SAE, ground-truth correlation on synthetic SAEs, and discriminability across training trajectories. We find that two of these metrics, Targeted Probe Perturbation (TPP) and Spurious Correlation Removal (SCR), fail multiple lenses at their canonical set","authors_text":"David Chanin","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-18T11:20:57Z","title":"Are Sparse Autoencoder Benchmarks Reliable?"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.18229","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:a7c694dc96f13546b2ed01c4ba8af4dc4da0464ba17570cde9556a040dfe9446","target":"record","created_at":"2026-05-20T00:05:51Z","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":"82bd1ecc62b69cf4b87140bdf005e06b458549039bf2966b9e4953efc9b33709","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-18T11:20:57Z","title_canon_sha256":"f3033cd9c6ce8b124b5a5289c0059d9d7579f3995274d3a9682ecbe0b24ced23"},"schema_version":"1.0","source":{"id":"2605.18229","kind":"arxiv","version":1}},"canonical_sha256":"34f3ce4e695db6cd628b3036c096ee730f9654436054cf37345194c0d9e8f057","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"34f3ce4e695db6cd628b3036c096ee730f9654436054cf37345194c0d9e8f057","first_computed_at":"2026-05-20T00:05:51.460020Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:05:51.460020Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"RkY4RnSTNdkuU53z72sIODJX7laiugsCC5u+KGs8LTSJ1taNEtA3a7tjPwHlSy5WWii7sNjJZ36RLqWUnoN3Cg==","signature_status":"signed_v1","signed_at":"2026-05-20T00:05:51.460441Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.18229","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:a7c694dc96f13546b2ed01c4ba8af4dc4da0464ba17570cde9556a040dfe9446","sha256:fd2758dca50324ee674a61e2bb0cc80ee94a5380944de89b3b28fd9853e94a20"],"state_sha256":"ecce3449b67c4eea7bce86e3e758b8f434645c8118b4287d21cec99ce5f69944"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"oscYt1LL3KMz6uqxGFkMTIpxs9AUBRIMztIVlr1sajT/f81y+NfrJKpm2pTpVBzwMcF3WRrsA3Hay6rW7xCfBw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-25T18:03:19.664280Z","bundle_sha256":"b7bf78f19fae683c1ac3f9b7005c6b59be96f4af3b5e4a570db76549f48f27c0"}}