{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XHEFZWLEIA7NC4NL4XWJGAA35K","short_pith_number":"pith:XHEFZWLE","schema_version":"1.0","canonical_sha256":"b9c85cd964403ed171abe5ec93001beaa6e54d8c071c8962b3950142f71cde49","source":{"kind":"arxiv","id":"2510.22228","version":2},"attestation_state":"computed","paper":{"title":"When Fewer Layers Break More Chains: Layer Pruning Harms Test-Time Scaling in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Guinan Su, Jonas Geiping, Keyu Wang, Lu Yin, Marco Canini, Shiwei Liu, Tian Lyu","submitted_at":"2025-10-25T09:22:22Z","abstract_excerpt":"Layer pruning has emerged as a widely adopted technique for improving the efficiency of large language models (LLMs). Although existing methods demonstrate strong performance retention on general knowledge tasks, their effect on long-chain reasoning, a more brittle yet crucial capability, remains largely unexplored. In this work, we study the impact of layer pruning on long-chain reasoning through the lens of test-time scaling, a key mechanism in modern LLMs that enables strong reasoning capacity by allocating more computation at inference time. With extensive experiments, we demonstrate that "},"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":"2510.22228","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-10-25T09:22:22Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"44d7d3bd2416fc43643483d52cb853a4622b8b246aedbfdc4a20b8dbcfa1df21","abstract_canon_sha256":"01963ad4537f83ee348a972241c463822ae557f04feb19f77277012c22904619"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-21T02:22:24.379964Z","signature_b64":"mgUEgREYqU0ZDFbmSV7eysjgnx7HR7vZaeLdnFJgfOM28IZc3OTI44XWNlCfDfQcVuVc7fPPBesikV3j5SXFCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b9c85cd964403ed171abe5ec93001beaa6e54d8c071c8962b3950142f71cde49","last_reissued_at":"2026-07-21T02:22:24.379114Z","signature_status":"signed_v1","first_computed_at":"2026-07-21T02:22:24.379114Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When Fewer Layers Break More Chains: Layer Pruning Harms Test-Time Scaling in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Guinan Su, Jonas Geiping, Keyu Wang, Lu Yin, Marco Canini, Shiwei Liu, Tian Lyu","submitted_at":"2025-10-25T09:22:22Z","abstract_excerpt":"Layer pruning has emerged as a widely adopted technique for improving the efficiency of large language models (LLMs). Although existing methods demonstrate strong performance retention on general knowledge tasks, their effect on long-chain reasoning, a more brittle yet crucial capability, remains largely unexplored. In this work, we study the impact of layer pruning on long-chain reasoning through the lens of test-time scaling, a key mechanism in modern LLMs that enables strong reasoning capacity by allocating more computation at inference time. With extensive experiments, we demonstrate that "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.22228","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/2510.22228/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":"2510.22228","created_at":"2026-07-21T02:22:24.379545+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.22228v2","created_at":"2026-07-21T02:22:24.379545+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.22228","created_at":"2026-07-21T02:22:24.379545+00:00"},{"alias_kind":"pith_short_12","alias_value":"XHEFZWLEIA7N","created_at":"2026-07-21T02:22:24.379545+00:00"},{"alias_kind":"pith_short_16","alias_value":"XHEFZWLEIA7NC4NL","created_at":"2026-07-21T02:22:24.379545+00:00"},{"alias_kind":"pith_short_8","alias_value":"XHEFZWLE","created_at":"2026-07-21T02:22:24.379545+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2512.22671","citing_title":"Fragile Knowledge, Robust Instruction-Following: The Width Pruning Dichotomy in Llama-3.2","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2602.01997","citing_title":"On the Limits of Layer Pruning for Generative Reasoning in Large Language Models","ref_index":30,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XHEFZWLEIA7NC4NL4XWJGAA35K","json":"https://pith.science/pith/XHEFZWLEIA7NC4NL4XWJGAA35K.json","graph_json":"https://pith.science/api/pith-number/XHEFZWLEIA7NC4NL4XWJGAA35K/graph.json","events_json":"https://pith.science/api/pith-number/XHEFZWLEIA7NC4NL4XWJGAA35K/events.json","paper":"https://pith.science/paper/XHEFZWLE"},"agent_actions":{"view_html":"https://pith.science/pith/XHEFZWLEIA7NC4NL4XWJGAA35K","download_json":"https://pith.science/pith/XHEFZWLEIA7NC4NL4XWJGAA35K.json","view_paper":"https://pith.science/paper/XHEFZWLE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.22228&json=true","fetch_graph":"https://pith.science/api/pith-number/XHEFZWLEIA7NC4NL4XWJGAA35K/graph.json","fetch_events":"https://pith.science/api/pith-number/XHEFZWLEIA7NC4NL4XWJGAA35K/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XHEFZWLEIA7NC4NL4XWJGAA35K/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XHEFZWLEIA7NC4NL4XWJGAA35K/action/storage_attestation","attest_author":"https://pith.science/pith/XHEFZWLEIA7NC4NL4XWJGAA35K/action/author_attestation","sign_citation":"https://pith.science/pith/XHEFZWLEIA7NC4NL4XWJGAA35K/action/citation_signature","submit_replication":"https://pith.science/pith/XHEFZWLEIA7NC4NL4XWJGAA35K/action/replication_record"}},"created_at":"2026-07-21T02:22:24.379545+00:00","updated_at":"2026-07-21T02:22:24.379545+00:00"}