{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XCMOLLDW2SRYYFAJCHR6OLK2MJ","short_pith_number":"pith:XCMOLLDW","schema_version":"1.0","canonical_sha256":"b898e5ac76d4a38c140911e3e72d5a6242e5c1462ac478d3a2005d8ef6381603","source":{"kind":"arxiv","id":"2505.14464","version":2},"attestation_state":"computed","paper":{"title":"Not All Correct Answers Are Equal: Why Your Distillation Source Matters","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Han Zhao, Haotian Wang, Shuaiting Chen, Sitong Zhao, Xiangang Li, Xiaoyu Tian, Yiping Peng, Yunjie Ji","submitted_at":"2025-05-20T15:00:51Z","abstract_excerpt":"Distillation has emerged as a practical and effective approach to enhance the reasoning capabilities of open-source language models. In this work, we conduct a large-scale empirical study on reasoning data distillation by collecting verified outputs from three state-of-the-art teacher models-AM-Thinking-v1, Qwen3-235B-A22B, and DeepSeek-R1-on a shared corpus of 1.89 million queries. We construct three parallel datasets and analyze their distributions, revealing that AM-Thinking-v1-distilled data exhibits greater token length diversity and lower perplexity. Student models trained on each datase"},"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":"2505.14464","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-05-20T15:00:51Z","cross_cats_sorted":[],"title_canon_sha256":"89b8e93377fe7a8419c4fd873c8b22300c5e79fba149f79b29dab49992761004","abstract_canon_sha256":"7e068235c356b10362c64499d11a8efbd5cdfa51ebc180291d799c6ac13d8a9a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:07:25.303014Z","signature_b64":"hgW7hHPjQH1yJO+anmGJlGGM6Le3w20M1W62hOus7Qe6As5JWUqSK5ToFakDDlxZeAWUtIMhE1ESKsyNBz0yBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b898e5ac76d4a38c140911e3e72d5a6242e5c1462ac478d3a2005d8ef6381603","last_reissued_at":"2026-07-05T11:07:25.302409Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:07:25.302409Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Not All Correct Answers Are Equal: Why Your Distillation Source Matters","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Han Zhao, Haotian Wang, Shuaiting Chen, Sitong Zhao, Xiangang Li, Xiaoyu Tian, Yiping Peng, Yunjie Ji","submitted_at":"2025-05-20T15:00:51Z","abstract_excerpt":"Distillation has emerged as a practical and effective approach to enhance the reasoning capabilities of open-source language models. In this work, we conduct a large-scale empirical study on reasoning data distillation by collecting verified outputs from three state-of-the-art teacher models-AM-Thinking-v1, Qwen3-235B-A22B, and DeepSeek-R1-on a shared corpus of 1.89 million queries. We construct three parallel datasets and analyze their distributions, revealing that AM-Thinking-v1-distilled data exhibits greater token length diversity and lower perplexity. Student models trained on each datase"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.14464","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/2505.14464/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":"2505.14464","created_at":"2026-07-05T11:07:25.302470+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.14464v2","created_at":"2026-07-05T11:07:25.302470+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.14464","created_at":"2026-07-05T11:07:25.302470+00:00"},{"alias_kind":"pith_short_12","alias_value":"XCMOLLDW2SRY","created_at":"2026-07-05T11:07:25.302470+00:00"},{"alias_kind":"pith_short_16","alias_value":"XCMOLLDW2SRYYFAJ","created_at":"2026-07-05T11:07:25.302470+00:00"},{"alias_kind":"pith_short_8","alias_value":"XCMOLLDW","created_at":"2026-07-05T11:07:25.302470+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26671","citing_title":"NebulaExp-8B: An Empirical Post-Training Pipeline via Full-Scale Ablation Research","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2511.11793","citing_title":"MiroThinker: Pushing the Performance Boundaries of Open-Source Research Agents via Model, Context, and Interactive Scaling","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10480","citing_title":"Tracing the Roots: A Multi-Agent Framework for Uncovering Data Lineage in Post-Training LLMs","ref_index":52,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XCMOLLDW2SRYYFAJCHR6OLK2MJ","json":"https://pith.science/pith/XCMOLLDW2SRYYFAJCHR6OLK2MJ.json","graph_json":"https://pith.science/api/pith-number/XCMOLLDW2SRYYFAJCHR6OLK2MJ/graph.json","events_json":"https://pith.science/api/pith-number/XCMOLLDW2SRYYFAJCHR6OLK2MJ/events.json","paper":"https://pith.science/paper/XCMOLLDW"},"agent_actions":{"view_html":"https://pith.science/pith/XCMOLLDW2SRYYFAJCHR6OLK2MJ","download_json":"https://pith.science/pith/XCMOLLDW2SRYYFAJCHR6OLK2MJ.json","view_paper":"https://pith.science/paper/XCMOLLDW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.14464&json=true","fetch_graph":"https://pith.science/api/pith-number/XCMOLLDW2SRYYFAJCHR6OLK2MJ/graph.json","fetch_events":"https://pith.science/api/pith-number/XCMOLLDW2SRYYFAJCHR6OLK2MJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XCMOLLDW2SRYYFAJCHR6OLK2MJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XCMOLLDW2SRYYFAJCHR6OLK2MJ/action/storage_attestation","attest_author":"https://pith.science/pith/XCMOLLDW2SRYYFAJCHR6OLK2MJ/action/author_attestation","sign_citation":"https://pith.science/pith/XCMOLLDW2SRYYFAJCHR6OLK2MJ/action/citation_signature","submit_replication":"https://pith.science/pith/XCMOLLDW2SRYYFAJCHR6OLK2MJ/action/replication_record"}},"created_at":"2026-07-05T11:07:25.302470+00:00","updated_at":"2026-07-05T11:07:25.302470+00:00"}