{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:UVKKH7MFGHQG22B7POPDPATBPP","short_pith_number":"pith:UVKKH7MF","schema_version":"1.0","canonical_sha256":"a554a3fd8531e06d683f7b9e3782617bf4a3ae086c52643a69fbcd8c0b697933","source":{"kind":"arxiv","id":"2402.14270","version":2},"attestation_state":"computed","paper":{"title":"Take the Bull by the Horns: Hard Sample-Reweighted Continual Training Improves LLM Generalization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Daouda Sow, Junjie Yang, Mingyuan Zhou, Tianlong Chen, Xuxi Chen, Yingbin Liang, Zhangyang Wang, Zhendong Wang","submitted_at":"2024-02-22T04:10:57Z","abstract_excerpt":"In the rapidly advancing arena of large language models (LLMs), a key challenge is to enhance their capabilities amid a looming shortage of high-quality training data. Our study starts from an empirical strategy for the light continual training of LLMs using their original pre-training data sets, with a specific focus on selective retention of samples that incur moderately high losses. These samples are deemed informative and beneficial for model refinement, contrasting with the highest-loss samples, which would be discarded due to their correlation with data noise and complexity. We then form"},"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":"2402.14270","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-22T04:10:57Z","cross_cats_sorted":[],"title_canon_sha256":"f95bb28957cc0cb08b299f7ea0c42c81321e759bf7442524fa1927f3688dc637","abstract_canon_sha256":"f9b02e508659b8a20edf7a162d484037f4a51d6e18a872b75b4cc46bdd7a2b90"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:50:55.848390Z","signature_b64":"6dRnpqOEJl/GnPySrH3Vs+0NuRt2piNfr6ytVs6aabak0Yk+D6VOJC/P+ejfufe8dWvxiOjCc4593jVhJazfAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a554a3fd8531e06d683f7b9e3782617bf4a3ae086c52643a69fbcd8c0b697933","last_reissued_at":"2026-07-05T07:50:55.847966Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:50:55.847966Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Take the Bull by the Horns: Hard Sample-Reweighted Continual Training Improves LLM Generalization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Daouda Sow, Junjie Yang, Mingyuan Zhou, Tianlong Chen, Xuxi Chen, Yingbin Liang, Zhangyang Wang, Zhendong Wang","submitted_at":"2024-02-22T04:10:57Z","abstract_excerpt":"In the rapidly advancing arena of large language models (LLMs), a key challenge is to enhance their capabilities amid a looming shortage of high-quality training data. Our study starts from an empirical strategy for the light continual training of LLMs using their original pre-training data sets, with a specific focus on selective retention of samples that incur moderately high losses. These samples are deemed informative and beneficial for model refinement, contrasting with the highest-loss samples, which would be discarded due to their correlation with data noise and complexity. We then form"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.14270","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/2402.14270/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":"2402.14270","created_at":"2026-07-05T07:50:55.848020+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.14270v2","created_at":"2026-07-05T07:50:55.848020+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.14270","created_at":"2026-07-05T07:50:55.848020+00:00"},{"alias_kind":"pith_short_12","alias_value":"UVKKH7MFGHQG","created_at":"2026-07-05T07:50:55.848020+00:00"},{"alias_kind":"pith_short_16","alias_value":"UVKKH7MFGHQG22B7","created_at":"2026-07-05T07:50:55.848020+00:00"},{"alias_kind":"pith_short_8","alias_value":"UVKKH7MF","created_at":"2026-07-05T07:50:55.848020+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2511.21692","citing_title":"Revisiting Generalization Across Difficulty Levels: It's Not So Easy","ref_index":6,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UVKKH7MFGHQG22B7POPDPATBPP","json":"https://pith.science/pith/UVKKH7MFGHQG22B7POPDPATBPP.json","graph_json":"https://pith.science/api/pith-number/UVKKH7MFGHQG22B7POPDPATBPP/graph.json","events_json":"https://pith.science/api/pith-number/UVKKH7MFGHQG22B7POPDPATBPP/events.json","paper":"https://pith.science/paper/UVKKH7MF"},"agent_actions":{"view_html":"https://pith.science/pith/UVKKH7MFGHQG22B7POPDPATBPP","download_json":"https://pith.science/pith/UVKKH7MFGHQG22B7POPDPATBPP.json","view_paper":"https://pith.science/paper/UVKKH7MF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.14270&json=true","fetch_graph":"https://pith.science/api/pith-number/UVKKH7MFGHQG22B7POPDPATBPP/graph.json","fetch_events":"https://pith.science/api/pith-number/UVKKH7MFGHQG22B7POPDPATBPP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UVKKH7MFGHQG22B7POPDPATBPP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UVKKH7MFGHQG22B7POPDPATBPP/action/storage_attestation","attest_author":"https://pith.science/pith/UVKKH7MFGHQG22B7POPDPATBPP/action/author_attestation","sign_citation":"https://pith.science/pith/UVKKH7MFGHQG22B7POPDPATBPP/action/citation_signature","submit_replication":"https://pith.science/pith/UVKKH7MFGHQG22B7POPDPATBPP/action/replication_record"}},"created_at":"2026-07-05T07:50:55.848020+00:00","updated_at":"2026-07-05T07:50:55.848020+00:00"}