{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XS62TQTR4HNR3FAM3EPLVX4V5I","short_pith_number":"pith:XS62TQTR","schema_version":"1.0","canonical_sha256":"bcbda9c271e1db1d940cd91ebadf95ea107a2572f66f148b396c1e3f232b49fa","source":{"kind":"arxiv","id":"2502.17717","version":1},"attestation_state":"computed","paper":{"title":"Knowledge Distillation with Training Wheels","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Abhinav Sethy, Anand Ramachandran, Guanlin Liu, Tanmay Gangwani, Yan Fu","submitted_at":"2025-02-24T23:17:52Z","abstract_excerpt":"Knowledge distillation is used, in generative language modeling, to train a smaller student model using the help of a larger teacher model, resulting in improved capabilities for the student model. In this paper, we formulate a more general framework for knowledge distillation where the student learns from the teacher during training, and also learns to ask for the teacher's help at test-time following rules specifying test-time restrictions. Towards this, we first formulate knowledge distillation as an entropy-regularized value optimization problem. Adopting Path Consistency Learning to solve"},"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":"2502.17717","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-24T23:17:52Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"f4899b78cc1a0d6aeebea1ba4af7b0f3c61efb3daa77dbc969e390751b63f56b","abstract_canon_sha256":"185ed232bcecb370c0bbc3995946270121360d962f123002070740ae6d89ba3b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:19:42.883243Z","signature_b64":"B0zLj76GwtadPpirzSUBhGtg08XQD9tEb7IOPUbOts9T5lBllMW08c8m+lm9oNESQwenNjiLsMajscNVvfl4Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bcbda9c271e1db1d940cd91ebadf95ea107a2572f66f148b396c1e3f232b49fa","last_reissued_at":"2026-07-05T10:19:42.882746Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:19:42.882746Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Knowledge Distillation with Training Wheels","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Abhinav Sethy, Anand Ramachandran, Guanlin Liu, Tanmay Gangwani, Yan Fu","submitted_at":"2025-02-24T23:17:52Z","abstract_excerpt":"Knowledge distillation is used, in generative language modeling, to train a smaller student model using the help of a larger teacher model, resulting in improved capabilities for the student model. In this paper, we formulate a more general framework for knowledge distillation where the student learns from the teacher during training, and also learns to ask for the teacher's help at test-time following rules specifying test-time restrictions. Towards this, we first formulate knowledge distillation as an entropy-regularized value optimization problem. Adopting Path Consistency Learning to solve"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.17717","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/2502.17717/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":"2502.17717","created_at":"2026-07-05T10:19:42.882805+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.17717v1","created_at":"2026-07-05T10:19:42.882805+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.17717","created_at":"2026-07-05T10:19:42.882805+00:00"},{"alias_kind":"pith_short_12","alias_value":"XS62TQTR4HNR","created_at":"2026-07-05T10:19:42.882805+00:00"},{"alias_kind":"pith_short_16","alias_value":"XS62TQTR4HNR3FAM","created_at":"2026-07-05T10:19:42.882805+00:00"},{"alias_kind":"pith_short_8","alias_value":"XS62TQTR","created_at":"2026-07-05T10:19:42.882805+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05339","citing_title":"TREK: Distill to Explore, Reinforce to Refine","ref_index":23,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12483","citing_title":"Beyond GRPO and On-Policy Distillation: An Empirical Sparse-to-Dense Reward Principle for Language-Model Post-Training","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12483","citing_title":"Beyond GRPO and On-Policy Distillation: An Empirical Sparse-to-Dense Reward Principle for Language-Model Post-Training","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12483","citing_title":"Beyond GRPO and On-Policy Distillation: An Empirical Sparse-to-Dense Reward Principle for Language-Model Post-Training","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12483","citing_title":"Beyond GRPO and On-Policy Distillation: An Empirical Sparse-to-Dense Reward Principle for Language-Model Post-Training","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XS62TQTR4HNR3FAM3EPLVX4V5I","json":"https://pith.science/pith/XS62TQTR4HNR3FAM3EPLVX4V5I.json","graph_json":"https://pith.science/api/pith-number/XS62TQTR4HNR3FAM3EPLVX4V5I/graph.json","events_json":"https://pith.science/api/pith-number/XS62TQTR4HNR3FAM3EPLVX4V5I/events.json","paper":"https://pith.science/paper/XS62TQTR"},"agent_actions":{"view_html":"https://pith.science/pith/XS62TQTR4HNR3FAM3EPLVX4V5I","download_json":"https://pith.science/pith/XS62TQTR4HNR3FAM3EPLVX4V5I.json","view_paper":"https://pith.science/paper/XS62TQTR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.17717&json=true","fetch_graph":"https://pith.science/api/pith-number/XS62TQTR4HNR3FAM3EPLVX4V5I/graph.json","fetch_events":"https://pith.science/api/pith-number/XS62TQTR4HNR3FAM3EPLVX4V5I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XS62TQTR4HNR3FAM3EPLVX4V5I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XS62TQTR4HNR3FAM3EPLVX4V5I/action/storage_attestation","attest_author":"https://pith.science/pith/XS62TQTR4HNR3FAM3EPLVX4V5I/action/author_attestation","sign_citation":"https://pith.science/pith/XS62TQTR4HNR3FAM3EPLVX4V5I/action/citation_signature","submit_replication":"https://pith.science/pith/XS62TQTR4HNR3FAM3EPLVX4V5I/action/replication_record"}},"created_at":"2026-07-05T10:19:42.882805+00:00","updated_at":"2026-07-05T10:19:42.882805+00:00"}