{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2021:URRXZ7SKKAT4VJ3QNXJBHGX4EV","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":"1fd99f798d5e3e1ea8271fe71cda43a44feeca748123c78c3a1324fcd6b44a5e","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-08-04T22:18:10Z","title_canon_sha256":"37a9fd7aa35f6d2e5854e35666c7566ef85b47c78478c7e4ee4e214fddfd39d0"},"schema_version":"1.0","source":{"id":"2108.07790","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2108.07790","created_at":"2026-07-05T03:35:31Z"},{"alias_kind":"arxiv_version","alias_value":"2108.07790v3","created_at":"2026-07-05T03:35:31Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2108.07790","created_at":"2026-07-05T03:35:31Z"},{"alias_kind":"pith_short_12","alias_value":"URRXZ7SKKAT4","created_at":"2026-07-05T03:35:31Z"},{"alias_kind":"pith_short_16","alias_value":"URRXZ7SKKAT4VJ3Q","created_at":"2026-07-05T03:35:31Z"},{"alias_kind":"pith_short_8","alias_value":"URRXZ7SK","created_at":"2026-07-05T03:35:31Z"}],"graph_snapshots":[{"event_id":"sha256:4b0b5e517ea65b63ae951da384ea0d7bfb6ac3ae39bea624fd6b1c8f5fd4dd2e","target":"graph","created_at":"2026-07-05T03:35:31Z","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/2108.07790/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Language models trained on large-scale unfiltered datasets curated from the open web acquire systemic biases, prejudices, and harmful views from their training data. We present a methodology for programmatically identifying and removing harmful text from web-scale datasets. A pretrained language model is used to calculate the log-likelihood of researcher-written trigger phrases conditioned on a specific document, which is used to identify and filter documents from the dataset. We demonstrate that models trained on this filtered dataset exhibit lower propensity to generate harmful text, with a ","authors_text":"Adrien Morisot, Carol Chen, Cooper Raterink, Helen Ngo, Ivan Zhang, Jo\\~ao G.M. Ara\\'ujo, Nicholas Frosst","cross_cats":["cs.LG"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-08-04T22:18:10Z","title":"Mitigating harm in language models with conditional-likelihood filtration"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2108.07790","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:b825e53aec7885e0752b12d263ea83f887636fcb592b5bd9fceda25f5e42d647","target":"record","created_at":"2026-07-05T03:35:31Z","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":"1fd99f798d5e3e1ea8271fe71cda43a44feeca748123c78c3a1324fcd6b44a5e","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-08-04T22:18:10Z","title_canon_sha256":"37a9fd7aa35f6d2e5854e35666c7566ef85b47c78478c7e4ee4e214fddfd39d0"},"schema_version":"1.0","source":{"id":"2108.07790","kind":"arxiv","version":3}},"canonical_sha256":"a4637cfe4a5027caa7706dd2139afc255b9c76730f2285b57823f7f3581c3db9","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a4637cfe4a5027caa7706dd2139afc255b9c76730f2285b57823f7f3581c3db9","first_computed_at":"2026-07-05T03:35:31.746474Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T03:35:31.746474Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"FrXvke+TJbNN9WoIYSFhB8KyIsVkQ6tTEC99c991MKsxdg/M4QX3EsCzE79Uf+zTU++0amfmUVA/UARwPlpvBg==","signature_status":"signed_v1","signed_at":"2026-07-05T03:35:31.746940Z","signed_message":"canonical_sha256_bytes"},"source_id":"2108.07790","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b825e53aec7885e0752b12d263ea83f887636fcb592b5bd9fceda25f5e42d647","sha256:4b0b5e517ea65b63ae951da384ea0d7bfb6ac3ae39bea624fd6b1c8f5fd4dd2e"],"state_sha256":"818d5c0d8ba2ee482fb8cdc18dccddfee857efd241c73670acc496c9833180b4"}