{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:2MUGHF4S4QTXU5NPKNOCAAJPYZ","short_pith_number":"pith:2MUGHF4S","schema_version":"1.0","canonical_sha256":"d328639792e4277a75af535c20012fc64853bd6779d72f5f07f6447e60f2f961","source":{"kind":"arxiv","id":"2606.19350","version":1},"attestation_state":"computed","paper":{"title":"Pruning via Causal Attribution Preserves Reasoning Performance in Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Amogh Sheth, Andrew Lin, Biruk Assefa, Yi Wen Huang, Yuhao Ge","submitted_at":"2026-04-27T01:44:10Z","abstract_excerpt":"Large language models (LLMs) excel at multi-step reasoning but incur substantial inference cost. We introduce Causal Attribution Pruning (CAP), a training-free method that identifies critical attention heads by measuring their causal impact on reasoning tasks and uses these head-level scores to guide fine-grained weight pruning. For each attention head, CAP estimates the expected performance degradation when the head is masked during forward passes on a small calibration set of reasoning problems. These causal scores are then converted into weight-level importance values for the corresponding "},"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":"2606.19350","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-27T01:44:10Z","cross_cats_sorted":[],"title_canon_sha256":"d2617f33259362177b4bec838886690c6a5158f628770b3ad93a2a22c444ba33","abstract_canon_sha256":"e98778b72c4d6abd76d8f5b9b36d432f86e7a0d54201b2305678ab580651b511"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:12:22.903630Z","signature_b64":"kirfMk3AipI2s/k+Ja8KZ+yqn7MTNkbVyBvtG/DNn1lG08ONleuRUaqgV/tqAmdww2sF2Mpd9A2+uWnj5xANCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d328639792e4277a75af535c20012fc64853bd6779d72f5f07f6447e60f2f961","last_reissued_at":"2026-06-19T16:12:22.903274Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:12:22.903274Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Pruning via Causal Attribution Preserves Reasoning Performance in Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Amogh Sheth, Andrew Lin, Biruk Assefa, Yi Wen Huang, Yuhao Ge","submitted_at":"2026-04-27T01:44:10Z","abstract_excerpt":"Large language models (LLMs) excel at multi-step reasoning but incur substantial inference cost. We introduce Causal Attribution Pruning (CAP), a training-free method that identifies critical attention heads by measuring their causal impact on reasoning tasks and uses these head-level scores to guide fine-grained weight pruning. For each attention head, CAP estimates the expected performance degradation when the head is masked during forward passes on a small calibration set of reasoning problems. These causal scores are then converted into weight-level importance values for the corresponding "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.19350","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/2606.19350/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":"2606.19350","created_at":"2026-06-19T16:12:22.903339+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.19350v1","created_at":"2026-06-19T16:12:22.903339+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.19350","created_at":"2026-06-19T16:12:22.903339+00:00"},{"alias_kind":"pith_short_12","alias_value":"2MUGHF4S4QTX","created_at":"2026-06-19T16:12:22.903339+00:00"},{"alias_kind":"pith_short_16","alias_value":"2MUGHF4S4QTXU5NP","created_at":"2026-06-19T16:12:22.903339+00:00"},{"alias_kind":"pith_short_8","alias_value":"2MUGHF4S","created_at":"2026-06-19T16:12:22.903339+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/2MUGHF4S4QTXU5NPKNOCAAJPYZ","json":"https://pith.science/pith/2MUGHF4S4QTXU5NPKNOCAAJPYZ.json","graph_json":"https://pith.science/api/pith-number/2MUGHF4S4QTXU5NPKNOCAAJPYZ/graph.json","events_json":"https://pith.science/api/pith-number/2MUGHF4S4QTXU5NPKNOCAAJPYZ/events.json","paper":"https://pith.science/paper/2MUGHF4S"},"agent_actions":{"view_html":"https://pith.science/pith/2MUGHF4S4QTXU5NPKNOCAAJPYZ","download_json":"https://pith.science/pith/2MUGHF4S4QTXU5NPKNOCAAJPYZ.json","view_paper":"https://pith.science/paper/2MUGHF4S","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.19350&json=true","fetch_graph":"https://pith.science/api/pith-number/2MUGHF4S4QTXU5NPKNOCAAJPYZ/graph.json","fetch_events":"https://pith.science/api/pith-number/2MUGHF4S4QTXU5NPKNOCAAJPYZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2MUGHF4S4QTXU5NPKNOCAAJPYZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2MUGHF4S4QTXU5NPKNOCAAJPYZ/action/storage_attestation","attest_author":"https://pith.science/pith/2MUGHF4S4QTXU5NPKNOCAAJPYZ/action/author_attestation","sign_citation":"https://pith.science/pith/2MUGHF4S4QTXU5NPKNOCAAJPYZ/action/citation_signature","submit_replication":"https://pith.science/pith/2MUGHF4S4QTXU5NPKNOCAAJPYZ/action/replication_record"}},"created_at":"2026-06-19T16:12:22.903339+00:00","updated_at":"2026-06-19T16:12:22.903339+00:00"}