{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:IXE7YJXU2JIGXETVHT3YHCUWTB","short_pith_number":"pith:IXE7YJXU","schema_version":"1.0","canonical_sha256":"45c9fc26f4d2506b92753cf7838a9698747d93d3d0a7e274e8a8051c166b9c3c","source":{"kind":"arxiv","id":"2506.04047","version":2},"attestation_state":"computed","paper":{"title":"On Support Samples of Next Word Prediction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Feifei Feng, Mou Xiao Feng, Yuanbin Wu, Yufang Liu, Yupei Du, Yuqian Li","submitted_at":"2025-06-04T15:13:22Z","abstract_excerpt":"Language models excel in various tasks by making complex decisions, yet understanding the rationale behind these decisions remains a challenge. This paper investigates \\emph{data-centric interpretability} in language models, focusing on the next-word prediction task. Using representer theorem, we identify two types of \\emph{support samples}-those that either promote or deter specific predictions. Our findings reveal that being a support sample is an intrinsic property, predictable even before training begins. Additionally, while non-support samples are less influential in direct predictions, t"},"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":"2506.04047","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-06-04T15:13:22Z","cross_cats_sorted":[],"title_canon_sha256":"ac7dc4dacc6776f9eb99d1709e3e64b49520aeb356318cb5e100cd6d5fcca63c","abstract_canon_sha256":"632ee5a74c64edb4ad81ab9a30f06d0440c7aecc6079e945e8fa5a218d0aa4b3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:18:22.926397Z","signature_b64":"7BP+Tu0tXPoKizo1tENLZupR6L3k+H3kjCj4C1wwtYCzOywveWGEzUARKNkeEfQCV0nHUb26UNXEybz7RtaoCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"45c9fc26f4d2506b92753cf7838a9698747d93d3d0a7e274e8a8051c166b9c3c","last_reissued_at":"2026-07-05T11:18:22.925878Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:18:22.925878Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On Support Samples of Next Word Prediction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Feifei Feng, Mou Xiao Feng, Yuanbin Wu, Yufang Liu, Yupei Du, Yuqian Li","submitted_at":"2025-06-04T15:13:22Z","abstract_excerpt":"Language models excel in various tasks by making complex decisions, yet understanding the rationale behind these decisions remains a challenge. This paper investigates \\emph{data-centric interpretability} in language models, focusing on the next-word prediction task. Using representer theorem, we identify two types of \\emph{support samples}-those that either promote or deter specific predictions. Our findings reveal that being a support sample is an intrinsic property, predictable even before training begins. Additionally, while non-support samples are less influential in direct predictions, t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.04047","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/2506.04047/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":"2506.04047","created_at":"2026-07-05T11:18:22.925938+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.04047v2","created_at":"2026-07-05T11:18:22.925938+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.04047","created_at":"2026-07-05T11:18:22.925938+00:00"},{"alias_kind":"pith_short_12","alias_value":"IXE7YJXU2JIG","created_at":"2026-07-05T11:18:22.925938+00:00"},{"alias_kind":"pith_short_16","alias_value":"IXE7YJXU2JIGXETV","created_at":"2026-07-05T11:18:22.925938+00:00"},{"alias_kind":"pith_short_8","alias_value":"IXE7YJXU","created_at":"2026-07-05T11:18:22.925938+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IXE7YJXU2JIGXETVHT3YHCUWTB","json":"https://pith.science/pith/IXE7YJXU2JIGXETVHT3YHCUWTB.json","graph_json":"https://pith.science/api/pith-number/IXE7YJXU2JIGXETVHT3YHCUWTB/graph.json","events_json":"https://pith.science/api/pith-number/IXE7YJXU2JIGXETVHT3YHCUWTB/events.json","paper":"https://pith.science/paper/IXE7YJXU"},"agent_actions":{"view_html":"https://pith.science/pith/IXE7YJXU2JIGXETVHT3YHCUWTB","download_json":"https://pith.science/pith/IXE7YJXU2JIGXETVHT3YHCUWTB.json","view_paper":"https://pith.science/paper/IXE7YJXU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.04047&json=true","fetch_graph":"https://pith.science/api/pith-number/IXE7YJXU2JIGXETVHT3YHCUWTB/graph.json","fetch_events":"https://pith.science/api/pith-number/IXE7YJXU2JIGXETVHT3YHCUWTB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IXE7YJXU2JIGXETVHT3YHCUWTB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IXE7YJXU2JIGXETVHT3YHCUWTB/action/storage_attestation","attest_author":"https://pith.science/pith/IXE7YJXU2JIGXETVHT3YHCUWTB/action/author_attestation","sign_citation":"https://pith.science/pith/IXE7YJXU2JIGXETVHT3YHCUWTB/action/citation_signature","submit_replication":"https://pith.science/pith/IXE7YJXU2JIGXETVHT3YHCUWTB/action/replication_record"}},"created_at":"2026-07-05T11:18:22.925938+00:00","updated_at":"2026-07-05T11:18:22.925938+00:00"}