{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:347D5IHBTPIV3NHYUQ227NOAWJ","short_pith_number":"pith:347D5IHB","schema_version":"1.0","canonical_sha256":"df3e3ea0e19bd15db4f8a435afb5c0b273e58af123874dce33ce23821472f6b8","source":{"kind":"arxiv","id":"2006.05656","version":4},"attestation_state":"computed","paper":{"title":"Why Attentions May Not Be Interpretable?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Bing Bai, Fei Wang, Guanhua Zhang, Hao Li, Jian Liang, Kun Bai","submitted_at":"2020-06-10T05:08:30Z","abstract_excerpt":"Attention-based methods have played important roles in model interpretations, where the calculated attention weights are expected to highlight the critical parts of inputs~(e.g., keywords in sentences). However, recent research found that attention-as-importance interpretations often do not work as we expected. For example, learned attention weights sometimes highlight less meaningful tokens like \"[SEP]\", \",\", and \".\", and are frequently uncorrelated with other feature importance indicators like gradient-based measures. A recent debate over whether attention is an explanation or not has drawn "},"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":"2006.05656","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2020-06-10T05:08:30Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"d3be3c63e753e2c9e5f678261f7a1d2755981273ab44959ea6ea1499e6938bc4","abstract_canon_sha256":"5af0e7391ab992dcfb12c98acb6af4c0e0884dab0c385520e63558eb6c100bb0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:45:48.024588Z","signature_b64":"QFZTLsxd1Jk2WrPi67L7SBOl/KwmBzRH5yFAkh7HwigqwhY+sYIQXNkxA4onaIEjx4Da5qcBruttGOxZxMn5BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"df3e3ea0e19bd15db4f8a435afb5c0b273e58af123874dce33ce23821472f6b8","last_reissued_at":"2026-07-05T02:45:48.024073Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:45:48.024073Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Why Attentions May Not Be Interpretable?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Bing Bai, Fei Wang, Guanhua Zhang, Hao Li, Jian Liang, Kun Bai","submitted_at":"2020-06-10T05:08:30Z","abstract_excerpt":"Attention-based methods have played important roles in model interpretations, where the calculated attention weights are expected to highlight the critical parts of inputs~(e.g., keywords in sentences). However, recent research found that attention-as-importance interpretations often do not work as we expected. For example, learned attention weights sometimes highlight less meaningful tokens like \"[SEP]\", \",\", and \".\", and are frequently uncorrelated with other feature importance indicators like gradient-based measures. A recent debate over whether attention is an explanation or not has drawn "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2006.05656","kind":"arxiv","version":4},"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/2006.05656/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":"2006.05656","created_at":"2026-07-05T02:45:48.024131+00:00"},{"alias_kind":"arxiv_version","alias_value":"2006.05656v4","created_at":"2026-07-05T02:45:48.024131+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2006.05656","created_at":"2026-07-05T02:45:48.024131+00:00"},{"alias_kind":"pith_short_12","alias_value":"347D5IHBTPIV","created_at":"2026-07-05T02:45:48.024131+00:00"},{"alias_kind":"pith_short_16","alias_value":"347D5IHBTPIV3NHY","created_at":"2026-07-05T02:45:48.024131+00:00"},{"alias_kind":"pith_short_8","alias_value":"347D5IHB","created_at":"2026-07-05T02:45:48.024131+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.23857","citing_title":"Emerging AI Approaches for Cancer Spatial Omics","ref_index":49,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/347D5IHBTPIV3NHYUQ227NOAWJ","json":"https://pith.science/pith/347D5IHBTPIV3NHYUQ227NOAWJ.json","graph_json":"https://pith.science/api/pith-number/347D5IHBTPIV3NHYUQ227NOAWJ/graph.json","events_json":"https://pith.science/api/pith-number/347D5IHBTPIV3NHYUQ227NOAWJ/events.json","paper":"https://pith.science/paper/347D5IHB"},"agent_actions":{"view_html":"https://pith.science/pith/347D5IHBTPIV3NHYUQ227NOAWJ","download_json":"https://pith.science/pith/347D5IHBTPIV3NHYUQ227NOAWJ.json","view_paper":"https://pith.science/paper/347D5IHB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2006.05656&json=true","fetch_graph":"https://pith.science/api/pith-number/347D5IHBTPIV3NHYUQ227NOAWJ/graph.json","fetch_events":"https://pith.science/api/pith-number/347D5IHBTPIV3NHYUQ227NOAWJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/347D5IHBTPIV3NHYUQ227NOAWJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/347D5IHBTPIV3NHYUQ227NOAWJ/action/storage_attestation","attest_author":"https://pith.science/pith/347D5IHBTPIV3NHYUQ227NOAWJ/action/author_attestation","sign_citation":"https://pith.science/pith/347D5IHBTPIV3NHYUQ227NOAWJ/action/citation_signature","submit_replication":"https://pith.science/pith/347D5IHBTPIV3NHYUQ227NOAWJ/action/replication_record"}},"created_at":"2026-07-05T02:45:48.024131+00:00","updated_at":"2026-07-05T02:45:48.024131+00:00"}