{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:ZBWLZWWXXZXIZPXSRFWPXRG3XW","short_pith_number":"pith:ZBWLZWWX","schema_version":"1.0","canonical_sha256":"c86cbcdad7be6e8cbef2896cfbc4dbbd8815a365ba60836f9c944b6059a1970c","source":{"kind":"arxiv","id":"2212.09095","version":2},"attestation_state":"computed","paper":{"title":"Rethinking the Role of Scale for In-Context Learning: An Interpretability-based Case Study at 66 Billion Scale","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Dan Roth, Hritik Bansal, Karthik Gopalakrishnan, Katrin Kirchhoff, Saket Dingliwal, Sravan Bodapati","submitted_at":"2022-12-18T14:36:07Z","abstract_excerpt":"Language models have been shown to perform better with an increase in scale on a wide variety of tasks via the in-context learning paradigm. In this paper, we investigate the hypothesis that the ability of a large language model to in-context learn-perform a task is not uniformly spread across all of its underlying components. Using a 66 billion parameter language model (OPT-66B) across a diverse set of 14 downstream tasks, we find this is indeed the case: $\\sim$70% of attention heads and $\\sim$20% of feed forward networks can be removed with minimal decline in task performance. We find substa"},"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":"2212.09095","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-12-18T14:36:07Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"83849c295c42726e45ad7a47533b5741f1ba2caf07c69826ca1aaf852ea9a805","abstract_canon_sha256":"2dbc94be03d2cc4cf17f98ead21738454bdf564a8ecadee51b602e5230065a52"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:41:46.572060Z","signature_b64":"0rpiX8KRJhivzrNO8mJW0nLpQ4zhlufzjB+f7H8hb5BB+CkqSXpBHaRxoAY8OkORJ6PWXkwNAK4k+sZ1ixSkBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c86cbcdad7be6e8cbef2896cfbc4dbbd8815a365ba60836f9c944b6059a1970c","last_reissued_at":"2026-07-05T06:41:46.571562Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:41:46.571562Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Rethinking the Role of Scale for In-Context Learning: An Interpretability-based Case Study at 66 Billion Scale","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Dan Roth, Hritik Bansal, Karthik Gopalakrishnan, Katrin Kirchhoff, Saket Dingliwal, Sravan Bodapati","submitted_at":"2022-12-18T14:36:07Z","abstract_excerpt":"Language models have been shown to perform better with an increase in scale on a wide variety of tasks via the in-context learning paradigm. In this paper, we investigate the hypothesis that the ability of a large language model to in-context learn-perform a task is not uniformly spread across all of its underlying components. Using a 66 billion parameter language model (OPT-66B) across a diverse set of 14 downstream tasks, we find this is indeed the case: $\\sim$70% of attention heads and $\\sim$20% of feed forward networks can be removed with minimal decline in task performance. We find substa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2212.09095","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/2212.09095/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":"2212.09095","created_at":"2026-07-05T06:41:46.571634+00:00"},{"alias_kind":"arxiv_version","alias_value":"2212.09095v2","created_at":"2026-07-05T06:41:46.571634+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2212.09095","created_at":"2026-07-05T06:41:46.571634+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZBWLZWWXXZXI","created_at":"2026-07-05T06:41:46.571634+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZBWLZWWXXZXIZPXS","created_at":"2026-07-05T06:41:46.571634+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZBWLZWWX","created_at":"2026-07-05T06:41:46.571634+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/ZBWLZWWXXZXIZPXSRFWPXRG3XW","json":"https://pith.science/pith/ZBWLZWWXXZXIZPXSRFWPXRG3XW.json","graph_json":"https://pith.science/api/pith-number/ZBWLZWWXXZXIZPXSRFWPXRG3XW/graph.json","events_json":"https://pith.science/api/pith-number/ZBWLZWWXXZXIZPXSRFWPXRG3XW/events.json","paper":"https://pith.science/paper/ZBWLZWWX"},"agent_actions":{"view_html":"https://pith.science/pith/ZBWLZWWXXZXIZPXSRFWPXRG3XW","download_json":"https://pith.science/pith/ZBWLZWWXXZXIZPXSRFWPXRG3XW.json","view_paper":"https://pith.science/paper/ZBWLZWWX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2212.09095&json=true","fetch_graph":"https://pith.science/api/pith-number/ZBWLZWWXXZXIZPXSRFWPXRG3XW/graph.json","fetch_events":"https://pith.science/api/pith-number/ZBWLZWWXXZXIZPXSRFWPXRG3XW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZBWLZWWXXZXIZPXSRFWPXRG3XW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZBWLZWWXXZXIZPXSRFWPXRG3XW/action/storage_attestation","attest_author":"https://pith.science/pith/ZBWLZWWXXZXIZPXSRFWPXRG3XW/action/author_attestation","sign_citation":"https://pith.science/pith/ZBWLZWWXXZXIZPXSRFWPXRG3XW/action/citation_signature","submit_replication":"https://pith.science/pith/ZBWLZWWXXZXIZPXSRFWPXRG3XW/action/replication_record"}},"created_at":"2026-07-05T06:41:46.571634+00:00","updated_at":"2026-07-05T06:41:46.571634+00:00"}