{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:RM3AUDYPPKO7JKVM7L32J7WIUC","short_pith_number":"pith:RM3AUDYP","schema_version":"1.0","canonical_sha256":"8b360a0f0f7a9df4aaacfaf7a4fec8a0b7de46e007c7dd1f4ae5cf63c94876c8","source":{"kind":"arxiv","id":"2504.05598","version":2},"attestation_state":"computed","paper":{"title":"DEL: Context-Aware Dynamic Exit Layer for Efficient Self-Speculative Decoding","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chaoyi Jiang, Hossein Entezari Zarch, Lei Gao, Murali Annavaram","submitted_at":"2025-04-08T01:12:59Z","abstract_excerpt":"Speculative Decoding (SD) is a widely used approach to accelerate the inference of large language models (LLMs) without reducing generation quality. It operates by first using a compact model to draft multiple tokens efficiently, followed by parallel verification using the target LLM. This approach leads to faster inference compared to auto-regressive decoding. While there are multiple approaches to create a draft model, one promising approach is to use early-exit methods. These methods draft candidate tokens by using a subset of layers of the primary model and applying the remaining layers fo"},"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":"2504.05598","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-08T01:12:59Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"ef85de7c8dbc2fc1cb94dc289f6e831d7e57e4228724bfca049cc3c52f6fad94","abstract_canon_sha256":"c1f65b1062244f373bd8556c12aaf92979a01598e201f76cceff64681ea57447"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:49:52.981175Z","signature_b64":"Ax16NJT+Qd+PKoWChFVWxAKHjb3qG/+O86jnWD7mSbFjok5XEjdVouYupVZon4khxGVP+qTzRAMTY8QFpTfKDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8b360a0f0f7a9df4aaacfaf7a4fec8a0b7de46e007c7dd1f4ae5cf63c94876c8","last_reissued_at":"2026-07-05T11:49:52.980704Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:49:52.980704Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DEL: Context-Aware Dynamic Exit Layer for Efficient Self-Speculative Decoding","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chaoyi Jiang, Hossein Entezari Zarch, Lei Gao, Murali Annavaram","submitted_at":"2025-04-08T01:12:59Z","abstract_excerpt":"Speculative Decoding (SD) is a widely used approach to accelerate the inference of large language models (LLMs) without reducing generation quality. It operates by first using a compact model to draft multiple tokens efficiently, followed by parallel verification using the target LLM. This approach leads to faster inference compared to auto-regressive decoding. While there are multiple approaches to create a draft model, one promising approach is to use early-exit methods. These methods draft candidate tokens by using a subset of layers of the primary model and applying the remaining layers fo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.05598","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/2504.05598/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":"2504.05598","created_at":"2026-07-05T11:49:52.980763+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.05598v2","created_at":"2026-07-05T11:49:52.980763+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.05598","created_at":"2026-07-05T11:49:52.980763+00:00"},{"alias_kind":"pith_short_12","alias_value":"RM3AUDYPPKO7","created_at":"2026-07-05T11:49:52.980763+00:00"},{"alias_kind":"pith_short_16","alias_value":"RM3AUDYPPKO7JKVM","created_at":"2026-07-05T11:49:52.980763+00:00"},{"alias_kind":"pith_short_8","alias_value":"RM3AUDYP","created_at":"2026-07-05T11:49:52.980763+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.01927","citing_title":"Scaling LLM Inference Beyond Amdahl`s Limits via Eliminating Non-Scalable Overheads","ref_index":54,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RM3AUDYPPKO7JKVM7L32J7WIUC","json":"https://pith.science/pith/RM3AUDYPPKO7JKVM7L32J7WIUC.json","graph_json":"https://pith.science/api/pith-number/RM3AUDYPPKO7JKVM7L32J7WIUC/graph.json","events_json":"https://pith.science/api/pith-number/RM3AUDYPPKO7JKVM7L32J7WIUC/events.json","paper":"https://pith.science/paper/RM3AUDYP"},"agent_actions":{"view_html":"https://pith.science/pith/RM3AUDYPPKO7JKVM7L32J7WIUC","download_json":"https://pith.science/pith/RM3AUDYPPKO7JKVM7L32J7WIUC.json","view_paper":"https://pith.science/paper/RM3AUDYP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.05598&json=true","fetch_graph":"https://pith.science/api/pith-number/RM3AUDYPPKO7JKVM7L32J7WIUC/graph.json","fetch_events":"https://pith.science/api/pith-number/RM3AUDYPPKO7JKVM7L32J7WIUC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RM3AUDYPPKO7JKVM7L32J7WIUC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RM3AUDYPPKO7JKVM7L32J7WIUC/action/storage_attestation","attest_author":"https://pith.science/pith/RM3AUDYPPKO7JKVM7L32J7WIUC/action/author_attestation","sign_citation":"https://pith.science/pith/RM3AUDYPPKO7JKVM7L32J7WIUC/action/citation_signature","submit_replication":"https://pith.science/pith/RM3AUDYPPKO7JKVM7L32J7WIUC/action/replication_record"}},"created_at":"2026-07-05T11:49:52.980763+00:00","updated_at":"2026-07-05T11:49:52.980763+00:00"}