{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:YK2STSPVNAJZRN6OSAVHNDL773","short_pith_number":"pith:YK2STSPV","schema_version":"1.0","canonical_sha256":"c2b529c9f5681398b7ce902a768d7ffef04f729989d6f6d2d753b9dc71b5b3db","source":{"kind":"arxiv","id":"2410.20022","version":1},"attestation_state":"computed","paper":{"title":"Dynamic layer selection in decoder-only transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Antonios Valkanas, Boris N. Oreshkin, Florence Regol, Joud Chataoui, Mark Coates, Theodore Glavas, Wassim Jabbour","submitted_at":"2024-10-26T00:44:11Z","abstract_excerpt":"The vast size of Large Language Models (LLMs) has prompted a search to optimize inference. One effective approach is dynamic inference, which adapts the architecture to the sample-at-hand to reduce the overall computational cost. We empirically examine two common dynamic inference methods for natural language generation (NLG): layer skipping and early exiting. We find that a pre-trained decoder-only model is significantly more robust to layer removal via layer skipping, as opposed to early exit. We demonstrate the difficulty of using hidden state information to adapt computation on a per-token"},"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":"2410.20022","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-10-26T00:44:11Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"3fe836821dd25c7700cb8153425c8870ac412f62a844971b013c9951fc500787","abstract_canon_sha256":"12cf495889b7db34c920bdc559628e7de8f139679bcee8fe86ae0e636d24cd1e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:26:37.689913Z","signature_b64":"OkEM42FCahBTI2xUbKyqWLeh8QZj6IPRFh8J0RfzWA3P+EBggeCx0w5jfp35EkgVXHWs1ZKmgjOlvmxY3JVJBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c2b529c9f5681398b7ce902a768d7ffef04f729989d6f6d2d753b9dc71b5b3db","last_reissued_at":"2026-07-05T09:26:37.689383Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:26:37.689383Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Dynamic layer selection in decoder-only transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Antonios Valkanas, Boris N. Oreshkin, Florence Regol, Joud Chataoui, Mark Coates, Theodore Glavas, Wassim Jabbour","submitted_at":"2024-10-26T00:44:11Z","abstract_excerpt":"The vast size of Large Language Models (LLMs) has prompted a search to optimize inference. One effective approach is dynamic inference, which adapts the architecture to the sample-at-hand to reduce the overall computational cost. We empirically examine two common dynamic inference methods for natural language generation (NLG): layer skipping and early exiting. We find that a pre-trained decoder-only model is significantly more robust to layer removal via layer skipping, as opposed to early exit. We demonstrate the difficulty of using hidden state information to adapt computation on a per-token"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.20022","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/2410.20022/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":"2410.20022","created_at":"2026-07-05T09:26:37.689449+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.20022v1","created_at":"2026-07-05T09:26:37.689449+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.20022","created_at":"2026-07-05T09:26:37.689449+00:00"},{"alias_kind":"pith_short_12","alias_value":"YK2STSPVNAJZ","created_at":"2026-07-05T09:26:37.689449+00:00"},{"alias_kind":"pith_short_16","alias_value":"YK2STSPVNAJZRN6O","created_at":"2026-07-05T09:26:37.689449+00:00"},{"alias_kind":"pith_short_8","alias_value":"YK2STSPV","created_at":"2026-07-05T09:26:37.689449+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/YK2STSPVNAJZRN6OSAVHNDL773","json":"https://pith.science/pith/YK2STSPVNAJZRN6OSAVHNDL773.json","graph_json":"https://pith.science/api/pith-number/YK2STSPVNAJZRN6OSAVHNDL773/graph.json","events_json":"https://pith.science/api/pith-number/YK2STSPVNAJZRN6OSAVHNDL773/events.json","paper":"https://pith.science/paper/YK2STSPV"},"agent_actions":{"view_html":"https://pith.science/pith/YK2STSPVNAJZRN6OSAVHNDL773","download_json":"https://pith.science/pith/YK2STSPVNAJZRN6OSAVHNDL773.json","view_paper":"https://pith.science/paper/YK2STSPV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.20022&json=true","fetch_graph":"https://pith.science/api/pith-number/YK2STSPVNAJZRN6OSAVHNDL773/graph.json","fetch_events":"https://pith.science/api/pith-number/YK2STSPVNAJZRN6OSAVHNDL773/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YK2STSPVNAJZRN6OSAVHNDL773/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YK2STSPVNAJZRN6OSAVHNDL773/action/storage_attestation","attest_author":"https://pith.science/pith/YK2STSPVNAJZRN6OSAVHNDL773/action/author_attestation","sign_citation":"https://pith.science/pith/YK2STSPVNAJZRN6OSAVHNDL773/action/citation_signature","submit_replication":"https://pith.science/pith/YK2STSPVNAJZRN6OSAVHNDL773/action/replication_record"}},"created_at":"2026-07-05T09:26:37.689449+00:00","updated_at":"2026-07-05T09:26:37.689449+00:00"}