{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:MKRS2OTCXP4T5YUQI7YT5CHD4W","short_pith_number":"pith:MKRS2OTC","schema_version":"1.0","canonical_sha256":"62a32d3a62bbf93ee29047f13e88e3e59747aaab06ff70bd0e56ae2f10e3cd88","source":{"kind":"arxiv","id":"2606.31796","version":1},"attestation_state":"computed","paper":{"title":"CHERRY: Compressed Hierarchical Experts with Recurrent Representational Yield","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Dohyeon Kwon, Youngjin Park","submitted_at":"2026-06-30T15:14:38Z","abstract_excerpt":"We study three complementary techniques for training compute-efficient language models.\n  (1) Selective supervision and per-token efficiency. Selective Ground Truth Token Training (SGT) concentrates supervision on the ~15% of output tokens that carry semantic payload. Through positive gradient coupling in position-shared transformer weights -- a token-level instance of auxiliary-task transfer -- the remaining 85% of unsupervised tokens still improve substantially, giving a 4.5x per-supervised-token efficiency (at the step-100 eval optimum, ~67% of the full-sequence loss reduction is recovered "},"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":"2606.31796","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-06-30T15:14:38Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2a05d2bf7defac00634a4be73cdffea1a4be8e5f9bc507625a9a812b69b090bd","abstract_canon_sha256":"72f037cb4f19484c456a5e22732455b3c2eac1187e535f64f5202522f4497492"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-01T01:18:15.508226Z","signature_b64":"m9W9cQG3OMINL2GwMyGkK59nJJYqcRqG5/cwTQ2BctRutSp9vtDLDzNoDEQsUKjPq0llGmgedRNgOuKW9tBKCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"62a32d3a62bbf93ee29047f13e88e3e59747aaab06ff70bd0e56ae2f10e3cd88","last_reissued_at":"2026-07-01T01:18:15.507710Z","signature_status":"signed_v1","first_computed_at":"2026-07-01T01:18:15.507710Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CHERRY: Compressed Hierarchical Experts with Recurrent Representational Yield","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Dohyeon Kwon, Youngjin Park","submitted_at":"2026-06-30T15:14:38Z","abstract_excerpt":"We study three complementary techniques for training compute-efficient language models.\n  (1) Selective supervision and per-token efficiency. Selective Ground Truth Token Training (SGT) concentrates supervision on the ~15% of output tokens that carry semantic payload. Through positive gradient coupling in position-shared transformer weights -- a token-level instance of auxiliary-task transfer -- the remaining 85% of unsupervised tokens still improve substantially, giving a 4.5x per-supervised-token efficiency (at the step-100 eval optimum, ~67% of the full-sequence loss reduction is recovered "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.31796","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/2606.31796/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":"2606.31796","created_at":"2026-07-01T01:18:15.507799+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.31796v1","created_at":"2026-07-01T01:18:15.507799+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.31796","created_at":"2026-07-01T01:18:15.507799+00:00"},{"alias_kind":"pith_short_12","alias_value":"MKRS2OTCXP4T","created_at":"2026-07-01T01:18:15.507799+00:00"},{"alias_kind":"pith_short_16","alias_value":"MKRS2OTCXP4T5YUQ","created_at":"2026-07-01T01:18:15.507799+00:00"},{"alias_kind":"pith_short_8","alias_value":"MKRS2OTC","created_at":"2026-07-01T01:18:15.507799+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/MKRS2OTCXP4T5YUQI7YT5CHD4W","json":"https://pith.science/pith/MKRS2OTCXP4T5YUQI7YT5CHD4W.json","graph_json":"https://pith.science/api/pith-number/MKRS2OTCXP4T5YUQI7YT5CHD4W/graph.json","events_json":"https://pith.science/api/pith-number/MKRS2OTCXP4T5YUQI7YT5CHD4W/events.json","paper":"https://pith.science/paper/MKRS2OTC"},"agent_actions":{"view_html":"https://pith.science/pith/MKRS2OTCXP4T5YUQI7YT5CHD4W","download_json":"https://pith.science/pith/MKRS2OTCXP4T5YUQI7YT5CHD4W.json","view_paper":"https://pith.science/paper/MKRS2OTC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.31796&json=true","fetch_graph":"https://pith.science/api/pith-number/MKRS2OTCXP4T5YUQI7YT5CHD4W/graph.json","fetch_events":"https://pith.science/api/pith-number/MKRS2OTCXP4T5YUQI7YT5CHD4W/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MKRS2OTCXP4T5YUQI7YT5CHD4W/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MKRS2OTCXP4T5YUQI7YT5CHD4W/action/storage_attestation","attest_author":"https://pith.science/pith/MKRS2OTCXP4T5YUQI7YT5CHD4W/action/author_attestation","sign_citation":"https://pith.science/pith/MKRS2OTCXP4T5YUQI7YT5CHD4W/action/citation_signature","submit_replication":"https://pith.science/pith/MKRS2OTCXP4T5YUQI7YT5CHD4W/action/replication_record"}},"created_at":"2026-07-01T01:18:15.507799+00:00","updated_at":"2026-07-01T01:18:15.507799+00:00"}