{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:4VVT6V4I73KTDJOGVSHVAWEQIQ","short_pith_number":"pith:4VVT6V4I","schema_version":"1.0","canonical_sha256":"e56b3f5788fed531a5c6ac8f50589044106850210315ff0931e4c2320d91bcc1","source":{"kind":"arxiv","id":"2605.22389","version":1},"attestation_state":"computed","paper":{"title":"Unified Data Selection for LLM Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chengpeng Li, Dayiheng Liu, Fengbin Zhu, Fuli Feng, Keqin Bao, Wenjie Wang, Xiaoyuan Li, Yiyao Yu, Yubo Ma","submitted_at":"2026-05-21T12:21:41Z","abstract_excerpt":"Effectively training Large Language Models (LLMs) for complex, long-CoT reasoning is often bottlenecked by the need for massive high-quality reasoning data. Existing methods are either computationally expensive or fail to reliably distinguish high- from low-quality reasoning samples. To address this, we propose High-Entropy Sum (HES), a training-free metric that quantifies reasoning quality by summing only the entropy of the top (e.g., 0.5\\%) highest-entropy tokens in each reasoning sample. We validate HES across three mainstream training paradigms: Supervised Fine-tuning (SFT), Rejection Fine"},"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":"2605.22389","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-21T12:21:41Z","cross_cats_sorted":[],"title_canon_sha256":"2022df57695539a08401510975b6e2e2afff74a226bd01cdb18da5a0bfbe9928","abstract_canon_sha256":"730816e7639183ff06fb2b29b392365d8b9d90cd7030f81801e9101e7d487631"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:04:41.166911Z","signature_b64":"rLvmrDmgo0QazLD5h0q3Bq5OvmM+UYoST4KwyKvm25kaJnCKcG82KlR58Fil6sFSmwIEGOlsheWAwdklyEdfCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e56b3f5788fed531a5c6ac8f50589044106850210315ff0931e4c2320d91bcc1","last_reissued_at":"2026-05-22T01:04:41.166088Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:04:41.166088Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unified Data Selection for LLM Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chengpeng Li, Dayiheng Liu, Fengbin Zhu, Fuli Feng, Keqin Bao, Wenjie Wang, Xiaoyuan Li, Yiyao Yu, Yubo Ma","submitted_at":"2026-05-21T12:21:41Z","abstract_excerpt":"Effectively training Large Language Models (LLMs) for complex, long-CoT reasoning is often bottlenecked by the need for massive high-quality reasoning data. Existing methods are either computationally expensive or fail to reliably distinguish high- from low-quality reasoning samples. To address this, we propose High-Entropy Sum (HES), a training-free metric that quantifies reasoning quality by summing only the entropy of the top (e.g., 0.5\\%) highest-entropy tokens in each reasoning sample. We validate HES across three mainstream training paradigms: Supervised Fine-tuning (SFT), Rejection Fine"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.22389","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/2605.22389/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":"2605.22389","created_at":"2026-05-22T01:04:41.166210+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.22389v1","created_at":"2026-05-22T01:04:41.166210+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.22389","created_at":"2026-05-22T01:04:41.166210+00:00"},{"alias_kind":"pith_short_12","alias_value":"4VVT6V4I73KT","created_at":"2026-05-22T01:04:41.166210+00:00"},{"alias_kind":"pith_short_16","alias_value":"4VVT6V4I73KTDJOG","created_at":"2026-05-22T01:04:41.166210+00:00"},{"alias_kind":"pith_short_8","alias_value":"4VVT6V4I","created_at":"2026-05-22T01:04:41.166210+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/4VVT6V4I73KTDJOGVSHVAWEQIQ","json":"https://pith.science/pith/4VVT6V4I73KTDJOGVSHVAWEQIQ.json","graph_json":"https://pith.science/api/pith-number/4VVT6V4I73KTDJOGVSHVAWEQIQ/graph.json","events_json":"https://pith.science/api/pith-number/4VVT6V4I73KTDJOGVSHVAWEQIQ/events.json","paper":"https://pith.science/paper/4VVT6V4I"},"agent_actions":{"view_html":"https://pith.science/pith/4VVT6V4I73KTDJOGVSHVAWEQIQ","download_json":"https://pith.science/pith/4VVT6V4I73KTDJOGVSHVAWEQIQ.json","view_paper":"https://pith.science/paper/4VVT6V4I","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.22389&json=true","fetch_graph":"https://pith.science/api/pith-number/4VVT6V4I73KTDJOGVSHVAWEQIQ/graph.json","fetch_events":"https://pith.science/api/pith-number/4VVT6V4I73KTDJOGVSHVAWEQIQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4VVT6V4I73KTDJOGVSHVAWEQIQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4VVT6V4I73KTDJOGVSHVAWEQIQ/action/storage_attestation","attest_author":"https://pith.science/pith/4VVT6V4I73KTDJOGVSHVAWEQIQ/action/author_attestation","sign_citation":"https://pith.science/pith/4VVT6V4I73KTDJOGVSHVAWEQIQ/action/citation_signature","submit_replication":"https://pith.science/pith/4VVT6V4I73KTDJOGVSHVAWEQIQ/action/replication_record"}},"created_at":"2026-05-22T01:04:41.166210+00:00","updated_at":"2026-05-22T01:04:41.166210+00:00"}