{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:I4UKAJXVHIEEGTES3NEPDLGTHQ","short_pith_number":"pith:I4UKAJXV","schema_version":"1.0","canonical_sha256":"4728a026f53a08434c92db48f1acd33c12433b374caf767091d97181e69a8698","source":{"kind":"arxiv","id":"2402.19465","version":2},"attestation_state":"computed","paper":{"title":"Towards Tracing Trustworthiness Dynamics: Revisiting Pre-training Period of Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chen Qian, Dongrui Liu, Jie Zhang, Jing Shao, Wei Yao, Yong Liu, Yu Qiao, Zhenfei Yin","submitted_at":"2024-02-29T18:55:06Z","abstract_excerpt":"Ensuring the trustworthiness of large language models (LLMs) is crucial. Most studies concentrate on fully pre-trained LLMs to better understand and improve LLMs' trustworthiness. In this paper, to reveal the untapped potential of pre-training, we pioneer the exploration of LLMs' trustworthiness during this period, focusing on five key dimensions: reliability, privacy, toxicity, fairness, and robustness. To begin with, we apply linear probing to LLMs. The high probing accuracy suggests that \\textit{LLMs in early pre-training can already distinguish concepts in each trustworthiness dimension}. "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2402.19465","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-29T18:55:06Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"46a0f2c939d5dd5160b3684ecaefec63a3493efb8edccabaad5de334f917d32e","abstract_canon_sha256":"6da5f211420fe1147bd5c91fa1ddd580dc51bb2e842531bd86d7f725d0bc8d29"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:01:30.530348Z","signature_b64":"a1Jv0xfo/m5biIhOD78b9Lofx6mQM9a4ddGljPSppEps8TsPs1Q4UCjYkNk0Ygh6ficqbeZKPkVFL8kqIDvNDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4728a026f53a08434c92db48f1acd33c12433b374caf767091d97181e69a8698","last_reissued_at":"2026-07-05T09:01:30.529866Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:01:30.529866Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Towards Tracing Trustworthiness Dynamics: Revisiting Pre-training Period of Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Chen Qian, Dongrui Liu, Jie Zhang, Jing Shao, Wei Yao, Yong Liu, Yu Qiao, Zhenfei Yin","submitted_at":"2024-02-29T18:55:06Z","abstract_excerpt":"Ensuring the trustworthiness of large language models (LLMs) is crucial. Most studies concentrate on fully pre-trained LLMs to better understand and improve LLMs' trustworthiness. In this paper, to reveal the untapped potential of pre-training, we pioneer the exploration of LLMs' trustworthiness during this period, focusing on five key dimensions: reliability, privacy, toxicity, fairness, and robustness. To begin with, we apply linear probing to LLMs. The high probing accuracy suggests that \\textit{LLMs in early pre-training can already distinguish concepts in each trustworthiness dimension}. "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.19465","kind":"arxiv","version":2},"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/2402.19465/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2402.19465","created_at":"2026-07-05T09:01:30.529923+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.19465v2","created_at":"2026-07-05T09:01:30.529923+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.19465","created_at":"2026-07-05T09:01:30.529923+00:00"},{"alias_kind":"pith_short_12","alias_value":"I4UKAJXVHIEE","created_at":"2026-07-05T09:01:30.529923+00:00"},{"alias_kind":"pith_short_16","alias_value":"I4UKAJXVHIEEGTES","created_at":"2026-07-05T09:01:30.529923+00:00"},{"alias_kind":"pith_short_8","alias_value":"I4UKAJXV","created_at":"2026-07-05T09:01:30.529923+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11375","citing_title":"When Probing Accuracy Saturates, Fragility Resolves: A Complementary Metric for LLM Pre-Training Analysis","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04572","citing_title":"From Parameter Dynamics to Risk Scoring : Quantifying Sample-Level Safety Degradation in LLM Fine-tuning","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/I4UKAJXVHIEEGTES3NEPDLGTHQ","json":"https://pith.science/pith/I4UKAJXVHIEEGTES3NEPDLGTHQ.json","graph_json":"https://pith.science/api/pith-number/I4UKAJXVHIEEGTES3NEPDLGTHQ/graph.json","events_json":"https://pith.science/api/pith-number/I4UKAJXVHIEEGTES3NEPDLGTHQ/events.json","paper":"https://pith.science/paper/I4UKAJXV"},"agent_actions":{"view_html":"https://pith.science/pith/I4UKAJXVHIEEGTES3NEPDLGTHQ","download_json":"https://pith.science/pith/I4UKAJXVHIEEGTES3NEPDLGTHQ.json","view_paper":"https://pith.science/paper/I4UKAJXV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.19465&json=true","fetch_graph":"https://pith.science/api/pith-number/I4UKAJXVHIEEGTES3NEPDLGTHQ/graph.json","fetch_events":"https://pith.science/api/pith-number/I4UKAJXVHIEEGTES3NEPDLGTHQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/I4UKAJXVHIEEGTES3NEPDLGTHQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/I4UKAJXVHIEEGTES3NEPDLGTHQ/action/storage_attestation","attest_author":"https://pith.science/pith/I4UKAJXVHIEEGTES3NEPDLGTHQ/action/author_attestation","sign_citation":"https://pith.science/pith/I4UKAJXVHIEEGTES3NEPDLGTHQ/action/citation_signature","submit_replication":"https://pith.science/pith/I4UKAJXVHIEEGTES3NEPDLGTHQ/action/replication_record"}},"created_at":"2026-07-05T09:01:30.529923+00:00","updated_at":"2026-07-05T09:01:30.529923+00:00"}