{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QJIYYNIFEKWA6I3K44K3HQ5Y6I","short_pith_number":"pith:QJIYYNIF","schema_version":"1.0","canonical_sha256":"82518c350522ac0f236ae715b3c3b8f234283d0d8dd1dcdbe338a31afc264e7a","source":{"kind":"arxiv","id":"2411.02344","version":2},"attestation_state":"computed","paper":{"title":"Seq-VCR: Preventing Collapse in Intermediate Transformer Representations for Enhanced Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Christopher Pal, Gopeshh Subbaraj, Irina Rish, Md Rifat Arefin, Nicolas Gontier, Ravid Shwartz-Ziv, Yann LeCun","submitted_at":"2024-11-04T18:14:07Z","abstract_excerpt":"Decoder-only Transformers often struggle with complex reasoning tasks, particularly arithmetic reasoning requiring multiple sequential operations. In this work, we identify representation collapse in the model's intermediate layers as a key factor limiting their reasoning capabilities. To address this, we propose Sequential Variance-Covariance Regularization (Seq-VCR), which enhances the entropy of intermediate representations and prevents collapse. Combined with dummy pause tokens as substitutes for chain-of-thought (CoT) tokens, our method significantly improves performance in arithmetic rea"},"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":"2411.02344","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-11-04T18:14:07Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"9c7a59da80081c8e0a5000a8defd2263dbd9dfeb2237b81ee7207df3d4d5b1d6","abstract_canon_sha256":"816b8676e99c6171051faa9d8a2c4ca086b032f465023efb195fa8a8778c371b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:36:15.267219Z","signature_b64":"EpRGBLX5r4cSIvvNJYsT4RbdmImErrgxpWsYU4ugWQ5iPLQi0M3Hyzrgv0Ab2+gfu7TGk3U0wuQ6lejEBRi8AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"82518c350522ac0f236ae715b3c3b8f234283d0d8dd1dcdbe338a31afc264e7a","last_reissued_at":"2026-07-05T10:36:15.266562Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:36:15.266562Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Seq-VCR: Preventing Collapse in Intermediate Transformer Representations for Enhanced Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Christopher Pal, Gopeshh Subbaraj, Irina Rish, Md Rifat Arefin, Nicolas Gontier, Ravid Shwartz-Ziv, Yann LeCun","submitted_at":"2024-11-04T18:14:07Z","abstract_excerpt":"Decoder-only Transformers often struggle with complex reasoning tasks, particularly arithmetic reasoning requiring multiple sequential operations. In this work, we identify representation collapse in the model's intermediate layers as a key factor limiting their reasoning capabilities. To address this, we propose Sequential Variance-Covariance Regularization (Seq-VCR), which enhances the entropy of intermediate representations and prevents collapse. Combined with dummy pause tokens as substitutes for chain-of-thought (CoT) tokens, our method significantly improves performance in arithmetic rea"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.02344","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/2411.02344/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":"2411.02344","created_at":"2026-07-05T10:36:15.266626+00:00"},{"alias_kind":"arxiv_version","alias_value":"2411.02344v2","created_at":"2026-07-05T10:36:15.266626+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.02344","created_at":"2026-07-05T10:36:15.266626+00:00"},{"alias_kind":"pith_short_12","alias_value":"QJIYYNIFEKWA","created_at":"2026-07-05T10:36:15.266626+00:00"},{"alias_kind":"pith_short_16","alias_value":"QJIYYNIFEKWA6I3K","created_at":"2026-07-05T10:36:15.266626+00:00"},{"alias_kind":"pith_short_8","alias_value":"QJIYYNIF","created_at":"2026-07-05T10:36:15.266626+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.17084","citing_title":"Scale Determines Whether Language Models Organize Representation Geometry for Prediction","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QJIYYNIFEKWA6I3K44K3HQ5Y6I","json":"https://pith.science/pith/QJIYYNIFEKWA6I3K44K3HQ5Y6I.json","graph_json":"https://pith.science/api/pith-number/QJIYYNIFEKWA6I3K44K3HQ5Y6I/graph.json","events_json":"https://pith.science/api/pith-number/QJIYYNIFEKWA6I3K44K3HQ5Y6I/events.json","paper":"https://pith.science/paper/QJIYYNIF"},"agent_actions":{"view_html":"https://pith.science/pith/QJIYYNIFEKWA6I3K44K3HQ5Y6I","download_json":"https://pith.science/pith/QJIYYNIFEKWA6I3K44K3HQ5Y6I.json","view_paper":"https://pith.science/paper/QJIYYNIF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2411.02344&json=true","fetch_graph":"https://pith.science/api/pith-number/QJIYYNIFEKWA6I3K44K3HQ5Y6I/graph.json","fetch_events":"https://pith.science/api/pith-number/QJIYYNIFEKWA6I3K44K3HQ5Y6I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QJIYYNIFEKWA6I3K44K3HQ5Y6I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QJIYYNIFEKWA6I3K44K3HQ5Y6I/action/storage_attestation","attest_author":"https://pith.science/pith/QJIYYNIFEKWA6I3K44K3HQ5Y6I/action/author_attestation","sign_citation":"https://pith.science/pith/QJIYYNIFEKWA6I3K44K3HQ5Y6I/action/citation_signature","submit_replication":"https://pith.science/pith/QJIYYNIFEKWA6I3K44K3HQ5Y6I/action/replication_record"}},"created_at":"2026-07-05T10:36:15.266626+00:00","updated_at":"2026-07-05T10:36:15.266626+00:00"}