{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:N6DZQHJ2ZXEIFJHYLSTLWBLABR","short_pith_number":"pith:N6DZQHJ2","schema_version":"1.0","canonical_sha256":"6f87981d3acdc882a4f85ca6bb05600c78c8cd39b868452e3f40af44bd3fe16a","source":{"kind":"arxiv","id":"2505.24244","version":1},"attestation_state":"computed","paper":{"title":"Mamba Knockout for Unraveling Factual Information Flow","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Idan Daniel Grosbard, Itay Tshuva, Nir Endy, Raja Giryes, Yonatan Slutzky, Yuval Ran-Milo","submitted_at":"2025-05-30T06:08:36Z","abstract_excerpt":"This paper investigates the flow of factual information in Mamba State-Space Model (SSM)-based language models. We rely on theoretical and empirical connections to Transformer-based architectures and their attention mechanisms. Exploiting this relationship, we adapt attentional interpretability techniques originally developed for Transformers--specifically, the Attention Knockout methodology--to both Mamba-1 and Mamba-2. Using them we trace how information is transmitted and localized across tokens and layers, revealing patterns of subject-token information emergence and layer-wise dynamics. N"},"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":"2505.24244","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-30T06:08:36Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"40a0ab7242876ec77050124b61310400efeeeafa97ae52fc0de517c32e778b35","abstract_canon_sha256":"fbf6fc72f70e31f34485f391bce7204e97394654bdcb8b3ad8976d93a9aab06e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:12:44.929393Z","signature_b64":"cIUaOLzOy/KaMuZNkMehfawn/MqRAkkGNb9ydT76cNVySY3Ygp+9RuZeuTslmxDaaBusJ+aPKYLKuKNmZBBXBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6f87981d3acdc882a4f85ca6bb05600c78c8cd39b868452e3f40af44bd3fe16a","last_reissued_at":"2026-07-05T11:12:44.928746Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:12:44.928746Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mamba Knockout for Unraveling Factual Information Flow","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Idan Daniel Grosbard, Itay Tshuva, Nir Endy, Raja Giryes, Yonatan Slutzky, Yuval Ran-Milo","submitted_at":"2025-05-30T06:08:36Z","abstract_excerpt":"This paper investigates the flow of factual information in Mamba State-Space Model (SSM)-based language models. We rely on theoretical and empirical connections to Transformer-based architectures and their attention mechanisms. Exploiting this relationship, we adapt attentional interpretability techniques originally developed for Transformers--specifically, the Attention Knockout methodology--to both Mamba-1 and Mamba-2. Using them we trace how information is transmitted and localized across tokens and layers, revealing patterns of subject-token information emergence and layer-wise dynamics. N"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.24244","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/2505.24244/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":"2505.24244","created_at":"2026-07-05T11:12:44.928854+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.24244v1","created_at":"2026-07-05T11:12:44.928854+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.24244","created_at":"2026-07-05T11:12:44.928854+00:00"},{"alias_kind":"pith_short_12","alias_value":"N6DZQHJ2ZXEI","created_at":"2026-07-05T11:12:44.928854+00:00"},{"alias_kind":"pith_short_16","alias_value":"N6DZQHJ2ZXEIFJHY","created_at":"2026-07-05T11:12:44.928854+00:00"},{"alias_kind":"pith_short_8","alias_value":"N6DZQHJ2","created_at":"2026-07-05T11:12:44.928854+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.12770","citing_title":"WriteSAE: Sparse Autoencoders for Recurrent State","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12770","citing_title":"WriteSAE: Sparse Autoencoders for Recurrent State","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12770","citing_title":"WriteSAE: Sparse Autoencoders for Recurrent State","ref_index":19,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/N6DZQHJ2ZXEIFJHYLSTLWBLABR","json":"https://pith.science/pith/N6DZQHJ2ZXEIFJHYLSTLWBLABR.json","graph_json":"https://pith.science/api/pith-number/N6DZQHJ2ZXEIFJHYLSTLWBLABR/graph.json","events_json":"https://pith.science/api/pith-number/N6DZQHJ2ZXEIFJHYLSTLWBLABR/events.json","paper":"https://pith.science/paper/N6DZQHJ2"},"agent_actions":{"view_html":"https://pith.science/pith/N6DZQHJ2ZXEIFJHYLSTLWBLABR","download_json":"https://pith.science/pith/N6DZQHJ2ZXEIFJHYLSTLWBLABR.json","view_paper":"https://pith.science/paper/N6DZQHJ2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.24244&json=true","fetch_graph":"https://pith.science/api/pith-number/N6DZQHJ2ZXEIFJHYLSTLWBLABR/graph.json","fetch_events":"https://pith.science/api/pith-number/N6DZQHJ2ZXEIFJHYLSTLWBLABR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/N6DZQHJ2ZXEIFJHYLSTLWBLABR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/N6DZQHJ2ZXEIFJHYLSTLWBLABR/action/storage_attestation","attest_author":"https://pith.science/pith/N6DZQHJ2ZXEIFJHYLSTLWBLABR/action/author_attestation","sign_citation":"https://pith.science/pith/N6DZQHJ2ZXEIFJHYLSTLWBLABR/action/citation_signature","submit_replication":"https://pith.science/pith/N6DZQHJ2ZXEIFJHYLSTLWBLABR/action/replication_record"}},"created_at":"2026-07-05T11:12:44.928854+00:00","updated_at":"2026-07-05T11:12:44.928854+00:00"}