{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:IDV7GTK75LIQMY3JJROKCUANCH","short_pith_number":"pith:IDV7GTK7","schema_version":"1.0","canonical_sha256":"40ebf34d5fead10663694c5ca1500d11e57b11909f4766b29b1a199e646fef2a","source":{"kind":"arxiv","id":"2507.01887","version":1},"attestation_state":"computed","paper":{"title":"MiCoTA: Bridging the Learnability Gap with Intermediate CoT and Teacher Assistants","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chenghao Zhu, Dongyi Ding, Meiling Tao, Tiannan Wang, Wangchunshu Zhou, Yuchen Eleanor Jiang","submitted_at":"2025-07-02T16:57:01Z","abstract_excerpt":"Large language models (LLMs) excel at reasoning tasks requiring long thought sequences for planning, reflection, and refinement. However, their substantial model size and high computational demands are impractical for widespread deployment. Yet, small language models (SLMs) often struggle to learn long-form CoT reasoning due to their limited capacity, a phenomenon we refer to as the \"SLMs Learnability Gap\". To address this, we introduce \\textbf{Mi}d-\\textbf{Co}T \\textbf{T}eacher \\textbf{A}ssistant Distillation (MiCoTAl), a framework for improving long CoT distillation for SLMs. MiCoTA employs "},"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":"2507.01887","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-07-02T16:57:01Z","cross_cats_sorted":[],"title_canon_sha256":"0471df53e9513b4f0094949af30dbebb8eea7e27d5a894f891c845b8a94d665f","abstract_canon_sha256":"9acf7b4d706ecf301060548958ccb08f351e8ddbaff18f26e49dd0dc97bef382"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:30:57.544507Z","signature_b64":"GvEF9J8A0rwHnvlVn4XUZ4v6ckRXY+v2P6dRnql2G2CFfS7JjLQ2/J91crdIlBHfX7dtO2dhdq2bgRvGDQGeDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"40ebf34d5fead10663694c5ca1500d11e57b11909f4766b29b1a199e646fef2a","last_reissued_at":"2026-07-05T11:30:57.543992Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:30:57.543992Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MiCoTA: Bridging the Learnability Gap with Intermediate CoT and Teacher Assistants","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chenghao Zhu, Dongyi Ding, Meiling Tao, Tiannan Wang, Wangchunshu Zhou, Yuchen Eleanor Jiang","submitted_at":"2025-07-02T16:57:01Z","abstract_excerpt":"Large language models (LLMs) excel at reasoning tasks requiring long thought sequences for planning, reflection, and refinement. However, their substantial model size and high computational demands are impractical for widespread deployment. Yet, small language models (SLMs) often struggle to learn long-form CoT reasoning due to their limited capacity, a phenomenon we refer to as the \"SLMs Learnability Gap\". To address this, we introduce \\textbf{Mi}d-\\textbf{Co}T \\textbf{T}eacher \\textbf{A}ssistant Distillation (MiCoTAl), a framework for improving long CoT distillation for SLMs. MiCoTA employs "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.01887","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/2507.01887/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":"2507.01887","created_at":"2026-07-05T11:30:57.544056+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.01887v1","created_at":"2026-07-05T11:30:57.544056+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.01887","created_at":"2026-07-05T11:30:57.544056+00:00"},{"alias_kind":"pith_short_12","alias_value":"IDV7GTK75LIQ","created_at":"2026-07-05T11:30:57.544056+00:00"},{"alias_kind":"pith_short_16","alias_value":"IDV7GTK75LIQMY3J","created_at":"2026-07-05T11:30:57.544056+00:00"},{"alias_kind":"pith_short_8","alias_value":"IDV7GTK7","created_at":"2026-07-05T11:30:57.544056+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.31159","citing_title":"Trust-Region Behavior Blending for On-Policy Distillation","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2601.14249","citing_title":"Which Reasoning Trajectories Teach Students to Reason Better? A Simple Metric of Informative Alignment","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IDV7GTK75LIQMY3JJROKCUANCH","json":"https://pith.science/pith/IDV7GTK75LIQMY3JJROKCUANCH.json","graph_json":"https://pith.science/api/pith-number/IDV7GTK75LIQMY3JJROKCUANCH/graph.json","events_json":"https://pith.science/api/pith-number/IDV7GTK75LIQMY3JJROKCUANCH/events.json","paper":"https://pith.science/paper/IDV7GTK7"},"agent_actions":{"view_html":"https://pith.science/pith/IDV7GTK75LIQMY3JJROKCUANCH","download_json":"https://pith.science/pith/IDV7GTK75LIQMY3JJROKCUANCH.json","view_paper":"https://pith.science/paper/IDV7GTK7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.01887&json=true","fetch_graph":"https://pith.science/api/pith-number/IDV7GTK75LIQMY3JJROKCUANCH/graph.json","fetch_events":"https://pith.science/api/pith-number/IDV7GTK75LIQMY3JJROKCUANCH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IDV7GTK75LIQMY3JJROKCUANCH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IDV7GTK75LIQMY3JJROKCUANCH/action/storage_attestation","attest_author":"https://pith.science/pith/IDV7GTK75LIQMY3JJROKCUANCH/action/author_attestation","sign_citation":"https://pith.science/pith/IDV7GTK75LIQMY3JJROKCUANCH/action/citation_signature","submit_replication":"https://pith.science/pith/IDV7GTK75LIQMY3JJROKCUANCH/action/replication_record"}},"created_at":"2026-07-05T11:30:57.544056+00:00","updated_at":"2026-07-05T11:30:57.544056+00:00"}