{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:R6CRE3WVTBRYJEJGX3EBLPAR3P","short_pith_number":"pith:R6CRE3WV","canonical_record":{"source":{"id":"2404.10464","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-16T11:07:48Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dd128dcc29e70fc2c4c53544385e43461e683b5ed48205ce71fc9a3cac926c01","abstract_canon_sha256":"f7411a882f74cac85f287c7a0524116308e8e1d87a2a16f467dcff62b439ed91"},"schema_version":"1.0"},"canonical_sha256":"8f85126ed59863849126bec815bc11dbe77c5c8950114caa49d89daeb4ed20b9","source":{"kind":"arxiv","id":"2404.10464","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2404.10464","created_at":"2026-07-05T08:54:10Z"},{"alias_kind":"arxiv_version","alias_value":"2404.10464v3","created_at":"2026-07-05T08:54:10Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.10464","created_at":"2026-07-05T08:54:10Z"},{"alias_kind":"pith_short_12","alias_value":"R6CRE3WVTBRY","created_at":"2026-07-05T08:54:10Z"},{"alias_kind":"pith_short_16","alias_value":"R6CRE3WVTBRYJEJG","created_at":"2026-07-05T08:54:10Z"},{"alias_kind":"pith_short_8","alias_value":"R6CRE3WV","created_at":"2026-07-05T08:54:10Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:R6CRE3WVTBRYJEJGX3EBLPAR3P","target":"record","payload":{"canonical_record":{"source":{"id":"2404.10464","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-16T11:07:48Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dd128dcc29e70fc2c4c53544385e43461e683b5ed48205ce71fc9a3cac926c01","abstract_canon_sha256":"f7411a882f74cac85f287c7a0524116308e8e1d87a2a16f467dcff62b439ed91"},"schema_version":"1.0"},"canonical_sha256":"8f85126ed59863849126bec815bc11dbe77c5c8950114caa49d89daeb4ed20b9","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:54:10.224521Z","signature_b64":"asax42mwoaTe8Y+fpBHWGgKYMXUfa9VRcNIMos01cvyoqSYrMB15I1xQk0B2wFXF/iU5k2vAA8qZ0MP67+6JCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8f85126ed59863849126bec815bc11dbe77c5c8950114caa49d89daeb4ed20b9","last_reissued_at":"2026-07-05T08:54:10.224081Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:54:10.224081Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2404.10464","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":"2026-07-05T08:54:10Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"AVG/r9qqxtCIsgtvyOlepqND1J6Bq2oGFLqpvpJ8ZfJKxRwFUY3slugdhAmaR9PwJ6D75ISn1/veotCdTeiUCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T11:18:44.769112Z"},"content_sha256":"086c0cc3b85649904c9d9ce79e228351e8f13e48f1b8344d862172f8364c1c0f","schema_version":"1.0","event_id":"sha256:086c0cc3b85649904c9d9ce79e228351e8f13e48f1b8344d862172f8364c1c0f"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:R6CRE3WVTBRYJEJGX3EBLPAR3P","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"DESTEIN: Navigating Detoxification of Language Models via Universal Steering Pairs and Head-wise Activation Fusion","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chuanyang Gong, Han Jiang, Yu Li, Zhihua Wei","submitted_at":"2024-04-16T11:07:48Z","abstract_excerpt":"Despite the remarkable achievements of language models (LMs) across a broad spectrum of tasks, their propensity for generating toxic outputs remains a prevalent concern. Current solutions involving finetuning or auxiliary models usually require extensive computational resources, hindering their practicality in large language models (LLMs). In this paper, we propose DeStein, a novel method that detoxifies LMs by applying representation engineering in activation spaces with lower resource and time costs. Specifically, we derive detoxification vectors from self-induced, universal steering pairs t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.10464","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/2404.10464/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":"2026-07-05T08:54:10Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6xxuy639PWr8IRzSGgIsib9O2ZsUhz+oDsOki8Stdi7w4RQMwA7EnOfuxYArNz4U4BlHHkLmJkkFJSnrZqXVBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T11:18:44.769750Z"},"content_sha256":"e61b7182d4847dd1cf65cb94081cef9ad64c3cd78ef31090657f454e2e0d7fb7","schema_version":"1.0","event_id":"sha256:e61b7182d4847dd1cf65cb94081cef9ad64c3cd78ef31090657f454e2e0d7fb7"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/R6CRE3WVTBRYJEJGX3EBLPAR3P/bundle.json","state_url":"https://pith.science/pith/R6CRE3WVTBRYJEJGX3EBLPAR3P/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/R6CRE3WVTBRYJEJGX3EBLPAR3P/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":"2026-08-05T11:18:44Z","links":{"resolver":"https://pith.science/pith/R6CRE3WVTBRYJEJGX3EBLPAR3P","bundle":"https://pith.science/pith/R6CRE3WVTBRYJEJGX3EBLPAR3P/bundle.json","state":"https://pith.science/pith/R6CRE3WVTBRYJEJGX3EBLPAR3P/state.json","well_known_bundle":"https://pith.science/.well-known/pith/R6CRE3WVTBRYJEJGX3EBLPAR3P/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:R6CRE3WVTBRYJEJGX3EBLPAR3P","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":"f7411a882f74cac85f287c7a0524116308e8e1d87a2a16f467dcff62b439ed91","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-16T11:07:48Z","title_canon_sha256":"dd128dcc29e70fc2c4c53544385e43461e683b5ed48205ce71fc9a3cac926c01"},"schema_version":"1.0","source":{"id":"2404.10464","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2404.10464","created_at":"2026-07-05T08:54:10Z"},{"alias_kind":"arxiv_version","alias_value":"2404.10464v3","created_at":"2026-07-05T08:54:10Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.10464","created_at":"2026-07-05T08:54:10Z"},{"alias_kind":"pith_short_12","alias_value":"R6CRE3WVTBRY","created_at":"2026-07-05T08:54:10Z"},{"alias_kind":"pith_short_16","alias_value":"R6CRE3WVTBRYJEJG","created_at":"2026-07-05T08:54:10Z"},{"alias_kind":"pith_short_8","alias_value":"R6CRE3WV","created_at":"2026-07-05T08:54:10Z"}],"graph_snapshots":[{"event_id":"sha256:e61b7182d4847dd1cf65cb94081cef9ad64c3cd78ef31090657f454e2e0d7fb7","target":"graph","created_at":"2026-07-05T08:54:10Z","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/2404.10464/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Despite the remarkable achievements of language models (LMs) across a broad spectrum of tasks, their propensity for generating toxic outputs remains a prevalent concern. Current solutions involving finetuning or auxiliary models usually require extensive computational resources, hindering their practicality in large language models (LLMs). In this paper, we propose DeStein, a novel method that detoxifies LMs by applying representation engineering in activation spaces with lower resource and time costs. Specifically, we derive detoxification vectors from self-induced, universal steering pairs t","authors_text":"Chuanyang Gong, Han Jiang, Yu Li, Zhihua Wei","cross_cats":["cs.AI"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-16T11:07:48Z","title":"DESTEIN: Navigating Detoxification of Language Models via Universal Steering Pairs and Head-wise Activation Fusion"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.10464","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":"sha256:086c0cc3b85649904c9d9ce79e228351e8f13e48f1b8344d862172f8364c1c0f","target":"record","created_at":"2026-07-05T08:54:10Z","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":"f7411a882f74cac85f287c7a0524116308e8e1d87a2a16f467dcff62b439ed91","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-16T11:07:48Z","title_canon_sha256":"dd128dcc29e70fc2c4c53544385e43461e683b5ed48205ce71fc9a3cac926c01"},"schema_version":"1.0","source":{"id":"2404.10464","kind":"arxiv","version":3}},"canonical_sha256":"8f85126ed59863849126bec815bc11dbe77c5c8950114caa49d89daeb4ed20b9","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"8f85126ed59863849126bec815bc11dbe77c5c8950114caa49d89daeb4ed20b9","first_computed_at":"2026-07-05T08:54:10.224081Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:54:10.224081Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"asax42mwoaTe8Y+fpBHWGgKYMXUfa9VRcNIMos01cvyoqSYrMB15I1xQk0B2wFXF/iU5k2vAA8qZ0MP67+6JCQ==","signature_status":"signed_v1","signed_at":"2026-07-05T08:54:10.224521Z","signed_message":"canonical_sha256_bytes"},"source_id":"2404.10464","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:086c0cc3b85649904c9d9ce79e228351e8f13e48f1b8344d862172f8364c1c0f","sha256:e61b7182d4847dd1cf65cb94081cef9ad64c3cd78ef31090657f454e2e0d7fb7"],"state_sha256":"74833e7ddceb8af3f4ce3fd06e0ff6f76c50ae3392df3070a62fb8f6ba1b598d"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"DihlHCbcTFT0JpRjldUUb/0+Nz8OMj4zqlKnIw9+z1R+4EfPoqHdHYzGwL5WBwGCQ+JsAXEBE0StQ/Lz+xaoDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-05T11:18:44.784805Z","bundle_sha256":"4e471783d9c0c4b0583d6a6e75af553d0a1b8b4214a6ea0de784595109620edb"}}