{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:XFDMH72Z24UHOLP6KWQMSNFZED","short_pith_number":"pith:XFDMH72Z","canonical_record":{"source":{"id":"2505.04741","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-07T19:17:49Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"025b44c66d3ce7023182a2adb700c9a2a28f3b7afd0191d2eb8ff43fef57fb95","abstract_canon_sha256":"a52cf697ac2b208e1db4c45d8f267ccf7e72b17f4c83163abbea522898b2750c"},"schema_version":"1.0"},"canonical_sha256":"b946c3ff59d728772dfe55a0c934b920f5171b82373dd5a7a71a210e174fe7fb","source":{"kind":"arxiv","id":"2505.04741","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.04741","created_at":"2026-07-05T11:00:13Z"},{"alias_kind":"arxiv_version","alias_value":"2505.04741v1","created_at":"2026-07-05T11:00:13Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.04741","created_at":"2026-07-05T11:00:13Z"},{"alias_kind":"pith_short_12","alias_value":"XFDMH72Z24UH","created_at":"2026-07-05T11:00:13Z"},{"alias_kind":"pith_short_16","alias_value":"XFDMH72Z24UHOLP6","created_at":"2026-07-05T11:00:13Z"},{"alias_kind":"pith_short_8","alias_value":"XFDMH72Z","created_at":"2026-07-05T11:00:13Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:XFDMH72Z24UHOLP6KWQMSNFZED","target":"record","payload":{"canonical_record":{"source":{"id":"2505.04741","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-07T19:17:49Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"025b44c66d3ce7023182a2adb700c9a2a28f3b7afd0191d2eb8ff43fef57fb95","abstract_canon_sha256":"a52cf697ac2b208e1db4c45d8f267ccf7e72b17f4c83163abbea522898b2750c"},"schema_version":"1.0"},"canonical_sha256":"b946c3ff59d728772dfe55a0c934b920f5171b82373dd5a7a71a210e174fe7fb","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:00:13.005434Z","signature_b64":"6BEK3gGVs02f0bWyLWAyhbdgVaCe8OuPw6e+hHRitejcY8ygsZTDF8FGNY/8BJFFf6nt19wE286oF1J2eDvICw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b946c3ff59d728772dfe55a0c934b920f5171b82373dd5a7a71a210e174fe7fb","last_reissued_at":"2026-07-05T11:00:13.004902Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:00:13.004902Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2505.04741","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":"2026-07-05T11:00:13Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"hJveCokHfE6w5vxd/TSms6FFMFWShjrAUXYJ+KYnfjxP4IxpYN0aiaLJAwoz8wnIX9oAeLE2xbv3HAau+ytQAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-15T11:23:31.015717Z"},"content_sha256":"b8c623f9322aa5d440452b648d2256876b5d24923f4ad5e9658c92ed86f3fe33","schema_version":"1.0","event_id":"sha256:b8c623f9322aa5d440452b648d2256876b5d24923f4ad5e9658c92ed86f3fe33"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:XFDMH72Z24UHOLP6KWQMSNFZED","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"When Bad Data Leads to Good Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Fernanda Vi\\'egas, Kenneth Li, Martin Wattenberg, Yida Chen","submitted_at":"2025-05-07T19:17:49Z","abstract_excerpt":"In large language model (LLM) pretraining, data quality is believed to determine model quality. In this paper, we re-examine the notion of \"quality\" from the perspective of pre- and post-training co-design. Specifically, we explore the possibility that pre-training on more toxic data can lead to better control in post-training, ultimately decreasing a model's output toxicity. First, we use a toy experiment to study how data composition affects the geometry of features in the representation space. Next, through controlled experiments with Olmo-1B models trained on varying ratios of clean and to"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.04741","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/2505.04741/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":"2026-07-05T11:00:13Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ynugwG/QKA1ivxeJ6Sj1BmyagkxvrNekZA8eAmsQzaVxC6vVhgd/2GFvwkAIv2NLSerqUesxGIHb7fIdhfOODg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-15T11:23:31.016222Z"},"content_sha256":"38676830e4952adc5cde13ee38ce850730a10a78b4421fe5bc35431808d7c678","schema_version":"1.0","event_id":"sha256:38676830e4952adc5cde13ee38ce850730a10a78b4421fe5bc35431808d7c678"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/XFDMH72Z24UHOLP6KWQMSNFZED/bundle.json","state_url":"https://pith.science/pith/XFDMH72Z24UHOLP6KWQMSNFZED/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/XFDMH72Z24UHOLP6KWQMSNFZED/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":"2026-08-15T11:23:31Z","links":{"resolver":"https://pith.science/pith/XFDMH72Z24UHOLP6KWQMSNFZED","bundle":"https://pith.science/pith/XFDMH72Z24UHOLP6KWQMSNFZED/bundle.json","state":"https://pith.science/pith/XFDMH72Z24UHOLP6KWQMSNFZED/state.json","well_known_bundle":"https://pith.science/.well-known/pith/XFDMH72Z24UHOLP6KWQMSNFZED/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:XFDMH72Z24UHOLP6KWQMSNFZED","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":"a52cf697ac2b208e1db4c45d8f267ccf7e72b17f4c83163abbea522898b2750c","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-07T19:17:49Z","title_canon_sha256":"025b44c66d3ce7023182a2adb700c9a2a28f3b7afd0191d2eb8ff43fef57fb95"},"schema_version":"1.0","source":{"id":"2505.04741","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.04741","created_at":"2026-07-05T11:00:13Z"},{"alias_kind":"arxiv_version","alias_value":"2505.04741v1","created_at":"2026-07-05T11:00:13Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.04741","created_at":"2026-07-05T11:00:13Z"},{"alias_kind":"pith_short_12","alias_value":"XFDMH72Z24UH","created_at":"2026-07-05T11:00:13Z"},{"alias_kind":"pith_short_16","alias_value":"XFDMH72Z24UHOLP6","created_at":"2026-07-05T11:00:13Z"},{"alias_kind":"pith_short_8","alias_value":"XFDMH72Z","created_at":"2026-07-05T11:00:13Z"}],"graph_snapshots":[{"event_id":"sha256:38676830e4952adc5cde13ee38ce850730a10a78b4421fe5bc35431808d7c678","target":"graph","created_at":"2026-07-05T11:00:13Z","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/2505.04741/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"In large language model (LLM) pretraining, data quality is believed to determine model quality. In this paper, we re-examine the notion of \"quality\" from the perspective of pre- and post-training co-design. Specifically, we explore the possibility that pre-training on more toxic data can lead to better control in post-training, ultimately decreasing a model's output toxicity. First, we use a toy experiment to study how data composition affects the geometry of features in the representation space. Next, through controlled experiments with Olmo-1B models trained on varying ratios of clean and to","authors_text":"Fernanda Vi\\'egas, Kenneth Li, Martin Wattenberg, Yida Chen","cross_cats":["cs.AI","cs.CL"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-07T19:17:49Z","title":"When Bad Data Leads to Good Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.04741","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:b8c623f9322aa5d440452b648d2256876b5d24923f4ad5e9658c92ed86f3fe33","target":"record","created_at":"2026-07-05T11:00:13Z","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":"a52cf697ac2b208e1db4c45d8f267ccf7e72b17f4c83163abbea522898b2750c","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-07T19:17:49Z","title_canon_sha256":"025b44c66d3ce7023182a2adb700c9a2a28f3b7afd0191d2eb8ff43fef57fb95"},"schema_version":"1.0","source":{"id":"2505.04741","kind":"arxiv","version":1}},"canonical_sha256":"b946c3ff59d728772dfe55a0c934b920f5171b82373dd5a7a71a210e174fe7fb","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"b946c3ff59d728772dfe55a0c934b920f5171b82373dd5a7a71a210e174fe7fb","first_computed_at":"2026-07-05T11:00:13.004902Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:00:13.004902Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"6BEK3gGVs02f0bWyLWAyhbdgVaCe8OuPw6e+hHRitejcY8ygsZTDF8FGNY/8BJFFf6nt19wE286oF1J2eDvICw==","signature_status":"signed_v1","signed_at":"2026-07-05T11:00:13.005434Z","signed_message":"canonical_sha256_bytes"},"source_id":"2505.04741","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b8c623f9322aa5d440452b648d2256876b5d24923f4ad5e9658c92ed86f3fe33","sha256:38676830e4952adc5cde13ee38ce850730a10a78b4421fe5bc35431808d7c678"],"state_sha256":"d9a261692b091a3dcdba3d81e1371e0479dddd5b487a2fc72e96a69778eba251"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1WcjRRInKaO6ESLR3gLNy4Czb+IjtzVD9PJlvprMohdLUUfdyBsGUqFFI6Dh7WIjtYHGTSDTbOgtuX5yVZD6DQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-15T11:23:31.019518Z","bundle_sha256":"1772d151e3476b7c00ef6c1e4a79a41f0c9f8d66dcc6bfda75011c224d5932a1"}}