{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:5YVXVX5F4ZDQQIBRAX3CRYWZW3","short_pith_number":"pith:5YVXVX5F","schema_version":"1.0","canonical_sha256":"ee2b7adfa5e64708203105f628e2d9b6d311a3380cc13a80ebc3fa29674b9df7","source":{"kind":"arxiv","id":"2312.13558","version":1},"attestation_state":"computed","paper":{"title":"The Truth is in There: Improving Reasoning in Language Models with Layer-Selective Rank Reduction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Dipendra Misra, Jordan T. Ash, Pratyusha Sharma","submitted_at":"2023-12-21T03:51:08Z","abstract_excerpt":"Transformer-based Large Language Models (LLMs) have become a fixture in modern machine learning. Correspondingly, significant resources are allocated towards research that aims to further advance this technology, typically resulting in models of increasing size that are trained on increasing amounts of data. This work, however, demonstrates the surprising result that it is often possible to significantly improve the performance of LLMs by selectively removing higher-order components of their weight matrices. This simple intervention, which we call LAyer-SElective Rank reduction (LASER), can be"},"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":"2312.13558","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-12-21T03:51:08Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CV"],"title_canon_sha256":"35e1d70d16a465abea3650ffa1269b06f99daa5b543e1e336ac853521b561687","abstract_canon_sha256":"5f2ee67bbf8e77f738f6c21a34e11e3e88737466c8aef4f05867b22e698cf46c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:26:46.549738Z","signature_b64":"yHcpMkz0IX37GClMwNoVqYUtQ/fatvqwABnda2axc1OTAW7Rz7tem89MMapnS5zXzFElqQyfN6xUqEKyTYH4Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ee2b7adfa5e64708203105f628e2d9b6d311a3380cc13a80ebc3fa29674b9df7","last_reissued_at":"2026-07-05T07:26:46.549337Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:26:46.549337Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Truth is in There: Improving Reasoning in Language Models with Layer-Selective Rank Reduction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Dipendra Misra, Jordan T. Ash, Pratyusha Sharma","submitted_at":"2023-12-21T03:51:08Z","abstract_excerpt":"Transformer-based Large Language Models (LLMs) have become a fixture in modern machine learning. Correspondingly, significant resources are allocated towards research that aims to further advance this technology, typically resulting in models of increasing size that are trained on increasing amounts of data. This work, however, demonstrates the surprising result that it is often possible to significantly improve the performance of LLMs by selectively removing higher-order components of their weight matrices. This simple intervention, which we call LAyer-SElective Rank reduction (LASER), can be"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2312.13558","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/2312.13558/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":"2312.13558","created_at":"2026-07-05T07:26:46.549395+00:00"},{"alias_kind":"arxiv_version","alias_value":"2312.13558v1","created_at":"2026-07-05T07:26:46.549395+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2312.13558","created_at":"2026-07-05T07:26:46.549395+00:00"},{"alias_kind":"pith_short_12","alias_value":"5YVXVX5F4ZDQ","created_at":"2026-07-05T07:26:46.549395+00:00"},{"alias_kind":"pith_short_16","alias_value":"5YVXVX5F4ZDQQIBR","created_at":"2026-07-05T07:26:46.549395+00:00"},{"alias_kind":"pith_short_8","alias_value":"5YVXVX5F","created_at":"2026-07-05T07:26:46.549395+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.23670","citing_title":"Tapered Language Models","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2603.10067","citing_title":"HTMuon: Improving Muon via Heavy-Tailed Spectral Correction","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2404.18923","citing_title":"Holmes: A Benchmark to Assess the Linguistic Competence of Language Models","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20296","citing_title":"Spectral Unforgetting: Post-Hoc Recovery of Damaged Capabilities Without Retraining","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16600","citing_title":"Where Pretraining writes and Alignment reads: the asymmetry of Transformer weight space","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15351","citing_title":"Aletheia: Gradient-Guided Layer Selection for Efficient LoRA Fine-Tuning Across Architectures","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01627","citing_title":"Importance-Guided Basis Selection for Low-Rank Decomposition of Large Language Models","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01609","citing_title":"Concepts Whisper While Syntax Shouts: Spectral Anti-Concentration and the Dual Geometry of Transformer Representations","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19351","citing_title":"DASH-KV: Accelerating Long-Context LLM Inference via Asymmetric KV Cache Hashing","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19974","citing_title":"Are LLM Uncertainty and Correctness Encoded by the Same Features? A Functional Dissociation via Sparse Autoencoders","ref_index":51,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5YVXVX5F4ZDQQIBRAX3CRYWZW3","json":"https://pith.science/pith/5YVXVX5F4ZDQQIBRAX3CRYWZW3.json","graph_json":"https://pith.science/api/pith-number/5YVXVX5F4ZDQQIBRAX3CRYWZW3/graph.json","events_json":"https://pith.science/api/pith-number/5YVXVX5F4ZDQQIBRAX3CRYWZW3/events.json","paper":"https://pith.science/paper/5YVXVX5F"},"agent_actions":{"view_html":"https://pith.science/pith/5YVXVX5F4ZDQQIBRAX3CRYWZW3","download_json":"https://pith.science/pith/5YVXVX5F4ZDQQIBRAX3CRYWZW3.json","view_paper":"https://pith.science/paper/5YVXVX5F","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2312.13558&json=true","fetch_graph":"https://pith.science/api/pith-number/5YVXVX5F4ZDQQIBRAX3CRYWZW3/graph.json","fetch_events":"https://pith.science/api/pith-number/5YVXVX5F4ZDQQIBRAX3CRYWZW3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5YVXVX5F4ZDQQIBRAX3CRYWZW3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5YVXVX5F4ZDQQIBRAX3CRYWZW3/action/storage_attestation","attest_author":"https://pith.science/pith/5YVXVX5F4ZDQQIBRAX3CRYWZW3/action/author_attestation","sign_citation":"https://pith.science/pith/5YVXVX5F4ZDQQIBRAX3CRYWZW3/action/citation_signature","submit_replication":"https://pith.science/pith/5YVXVX5F4ZDQQIBRAX3CRYWZW3/action/replication_record"}},"created_at":"2026-07-05T07:26:46.549395+00:00","updated_at":"2026-07-05T07:26:46.549395+00:00"}