{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:W3P5NA7EKRJFF2RPI3MLTIZ3QM","short_pith_number":"pith:W3P5NA7E","canonical_record":{"source":{"id":"2406.04313","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-06T17:57:04Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CV","cs.CY"],"title_canon_sha256":"c04f56e050bcf64b0372c0d09fe38d0db9a8d4c423ce1bf22a37607396f0b6e8","abstract_canon_sha256":"86006e74193841e489621a3299bf21bd9ecc614904928aef9eb470c42fcdf7e6"},"schema_version":"1.0"},"canonical_sha256":"b6dfd683e4545252ea2f46d8b9a33b8316b9f17d2d91adae576e75101a900287","source":{"kind":"arxiv","id":"2406.04313","version":4},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.04313","created_at":"2026-07-05T08:43:10Z"},{"alias_kind":"arxiv_version","alias_value":"2406.04313v4","created_at":"2026-07-05T08:43:10Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.04313","created_at":"2026-07-05T08:43:10Z"},{"alias_kind":"pith_short_12","alias_value":"W3P5NA7EKRJF","created_at":"2026-07-05T08:43:10Z"},{"alias_kind":"pith_short_16","alias_value":"W3P5NA7EKRJFF2RP","created_at":"2026-07-05T08:43:10Z"},{"alias_kind":"pith_short_8","alias_value":"W3P5NA7E","created_at":"2026-07-05T08:43:10Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:W3P5NA7EKRJFF2RPI3MLTIZ3QM","target":"record","payload":{"canonical_record":{"source":{"id":"2406.04313","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-06T17:57:04Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CV","cs.CY"],"title_canon_sha256":"c04f56e050bcf64b0372c0d09fe38d0db9a8d4c423ce1bf22a37607396f0b6e8","abstract_canon_sha256":"86006e74193841e489621a3299bf21bd9ecc614904928aef9eb470c42fcdf7e6"},"schema_version":"1.0"},"canonical_sha256":"b6dfd683e4545252ea2f46d8b9a33b8316b9f17d2d91adae576e75101a900287","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:43:10.553271Z","signature_b64":"mXwCdtOqNL5Xw5WIEYso0mvFLJ4britN/lpZgUcjQBC8b6p+nFfDc0TiBgagib5HDKoScjiHGdPWMeJUtYlWDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b6dfd683e4545252ea2f46d8b9a33b8316b9f17d2d91adae576e75101a900287","last_reissued_at":"2026-07-05T08:43:10.552792Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:43:10.552792Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2406.04313","source_version":4,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:43:10Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"uHUFsvC4a3rMYe4qw0kr6hS03QZeX9mej2362mIeBpo00psPJDJJYO5WcdQt7iMAuEjDkXU0G+ZZ5+p0BUzkAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T10:01:39.675098Z"},"content_sha256":"29865f02b0f579d408ed215bc472ae62fd435ab46d4470fe295c133ba725348a","schema_version":"1.0","event_id":"sha256:29865f02b0f579d408ed215bc472ae62fd435ab46d4470fe295c133ba725348a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:W3P5NA7EKRJFF2RPI3MLTIZ3QM","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Improving Alignment and Robustness with Circuit Breakers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CV","cs.CY"],"primary_cat":"cs.LG","authors_text":"Andy Zou, Dan Hendrycks, Derek Duenas, Justin Wang, Long Phan, Maksym Andriushchenko, Matt Fredrikson, Maxwell Lin, Rowan Wang, Zico Kolter","submitted_at":"2024-06-06T17:57:04Z","abstract_excerpt":"AI systems can take harmful actions and are highly vulnerable to adversarial attacks. We present an approach, inspired by recent advances in representation engineering, that interrupts the models as they respond with harmful outputs with \"circuit breakers.\" Existing techniques aimed at improving alignment, such as refusal training, are often bypassed. Techniques such as adversarial training try to plug these holes by countering specific attacks. As an alternative to refusal training and adversarial training, circuit-breaking directly controls the representations that are responsible for harmfu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.04313","kind":"arxiv","version":4},"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/2406.04313/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":"2026-07-05T08:43:10Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"xo8rQ20dRBM+G96HAyIuEpSDPNPySrXtiyQjumrEPQdJht5AF6/W6jsjTtz1csgKDGh7l+elYarb8vOLgvxUDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T10:01:39.675483Z"},"content_sha256":"fbd2242085fdcb1585f8dc536245714b08d7405c655ae2ff2c09c5eafdca24bf","schema_version":"1.0","event_id":"sha256:fbd2242085fdcb1585f8dc536245714b08d7405c655ae2ff2c09c5eafdca24bf"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/W3P5NA7EKRJFF2RPI3MLTIZ3QM/bundle.json","state_url":"https://pith.science/pith/W3P5NA7EKRJFF2RPI3MLTIZ3QM/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/W3P5NA7EKRJFF2RPI3MLTIZ3QM/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":"2026-07-07T10:01:39Z","links":{"resolver":"https://pith.science/pith/W3P5NA7EKRJFF2RPI3MLTIZ3QM","bundle":"https://pith.science/pith/W3P5NA7EKRJFF2RPI3MLTIZ3QM/bundle.json","state":"https://pith.science/pith/W3P5NA7EKRJFF2RPI3MLTIZ3QM/state.json","well_known_bundle":"https://pith.science/.well-known/pith/W3P5NA7EKRJFF2RPI3MLTIZ3QM/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:W3P5NA7EKRJFF2RPI3MLTIZ3QM","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":"86006e74193841e489621a3299bf21bd9ecc614904928aef9eb470c42fcdf7e6","cross_cats_sorted":["cs.AI","cs.CL","cs.CV","cs.CY"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-06T17:57:04Z","title_canon_sha256":"c04f56e050bcf64b0372c0d09fe38d0db9a8d4c423ce1bf22a37607396f0b6e8"},"schema_version":"1.0","source":{"id":"2406.04313","kind":"arxiv","version":4}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.04313","created_at":"2026-07-05T08:43:10Z"},{"alias_kind":"arxiv_version","alias_value":"2406.04313v4","created_at":"2026-07-05T08:43:10Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.04313","created_at":"2026-07-05T08:43:10Z"},{"alias_kind":"pith_short_12","alias_value":"W3P5NA7EKRJF","created_at":"2026-07-05T08:43:10Z"},{"alias_kind":"pith_short_16","alias_value":"W3P5NA7EKRJFF2RP","created_at":"2026-07-05T08:43:10Z"},{"alias_kind":"pith_short_8","alias_value":"W3P5NA7E","created_at":"2026-07-05T08:43:10Z"}],"graph_snapshots":[{"event_id":"sha256:fbd2242085fdcb1585f8dc536245714b08d7405c655ae2ff2c09c5eafdca24bf","target":"graph","created_at":"2026-07-05T08:43:10Z","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/2406.04313/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"AI systems can take harmful actions and are highly vulnerable to adversarial attacks. We present an approach, inspired by recent advances in representation engineering, that interrupts the models as they respond with harmful outputs with \"circuit breakers.\" Existing techniques aimed at improving alignment, such as refusal training, are often bypassed. Techniques such as adversarial training try to plug these holes by countering specific attacks. As an alternative to refusal training and adversarial training, circuit-breaking directly controls the representations that are responsible for harmfu","authors_text":"Andy Zou, Dan Hendrycks, Derek Duenas, Justin Wang, Long Phan, Maksym Andriushchenko, Matt Fredrikson, Maxwell Lin, Rowan Wang, Zico Kolter","cross_cats":["cs.AI","cs.CL","cs.CV","cs.CY"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-06T17:57:04Z","title":"Improving Alignment and Robustness with Circuit Breakers"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.04313","kind":"arxiv","version":4},"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":"sha256:29865f02b0f579d408ed215bc472ae62fd435ab46d4470fe295c133ba725348a","target":"record","created_at":"2026-07-05T08:43:10Z","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":"86006e74193841e489621a3299bf21bd9ecc614904928aef9eb470c42fcdf7e6","cross_cats_sorted":["cs.AI","cs.CL","cs.CV","cs.CY"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-06T17:57:04Z","title_canon_sha256":"c04f56e050bcf64b0372c0d09fe38d0db9a8d4c423ce1bf22a37607396f0b6e8"},"schema_version":"1.0","source":{"id":"2406.04313","kind":"arxiv","version":4}},"canonical_sha256":"b6dfd683e4545252ea2f46d8b9a33b8316b9f17d2d91adae576e75101a900287","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"b6dfd683e4545252ea2f46d8b9a33b8316b9f17d2d91adae576e75101a900287","first_computed_at":"2026-07-05T08:43:10.552792Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:43:10.552792Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"mXwCdtOqNL5Xw5WIEYso0mvFLJ4britN/lpZgUcjQBC8b6p+nFfDc0TiBgagib5HDKoScjiHGdPWMeJUtYlWDw==","signature_status":"signed_v1","signed_at":"2026-07-05T08:43:10.553271Z","signed_message":"canonical_sha256_bytes"},"source_id":"2406.04313","source_kind":"arxiv","source_version":4}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:29865f02b0f579d408ed215bc472ae62fd435ab46d4470fe295c133ba725348a","sha256:fbd2242085fdcb1585f8dc536245714b08d7405c655ae2ff2c09c5eafdca24bf"],"state_sha256":"5f33b47d6299f753983b300dc35fb561382eb8180cca37d56994ce7576e470f4"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"LS2a083WLH1TzYCBa8bqwnbI3IYhfnLD4YC3dCtbc9pxoAU6yY3426ZNmpjyRuIgQn8GAWe8OtHxf/PIW3OLAA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-07T10:01:39.677393Z","bundle_sha256":"e872b840baff1fc21de136190e9a366055c22a5c1d717939f74fe77f77af1c2f"}}