{"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"}