{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:KXX4LGLP52GMCPZ7SIK7PD35RD","short_pith_number":"pith:KXX4LGLP","schema_version":"1.0","canonical_sha256":"55efc5996fee8cc13f3f9215f78f7d88ea654a31b0d7139d5858a9ae168f2ed6","source":{"kind":"arxiv","id":"2406.16858","version":2},"attestation_state":"computed","paper":{"title":"EAGLE-2: Faster Inference of Language Models with Dynamic Draft Trees","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chao Zhang, Fangyun Wei, Hongyang Zhang, Yuhui Li","submitted_at":"2024-06-24T17:59:11Z","abstract_excerpt":"Inference with modern Large Language Models (LLMs) is expensive and time-consuming, and speculative sampling has proven to be an effective solution. Most speculative sampling methods such as EAGLE use a static draft tree, implicitly assuming that the acceptance rate of draft tokens depends only on their position. Interestingly, we found that the acceptance rate of draft tokens is also context-dependent. In this paper, building upon EAGLE, we propose EAGLE-2, which introduces a new technique of context-aware dynamic draft tree into drafting modeling. This improvement leverages the fact that the"},"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":"2406.16858","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-06-24T17:59:11Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"2ffc0f52cde11f3cf895871ddced219d62f50d958e7f387bd652b093fa71441b","abstract_canon_sha256":"4615e7454dd1abb94ae8327ae9d1bc3be7e393dead6d81d60a30aa992f467811"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:38:11.848092Z","signature_b64":"KZGw9OwDvna/UwBgXqwbx9G3y/p+/vPaSkZJoT2Dy4fNUiFOXYmUPPzXCNNRI3oyek56V3F1ErxkyMcMbPIOBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"55efc5996fee8cc13f3f9215f78f7d88ea654a31b0d7139d5858a9ae168f2ed6","last_reissued_at":"2026-07-05T08:38:11.847582Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:38:11.847582Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"EAGLE-2: Faster Inference of Language Models with Dynamic Draft Trees","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Chao Zhang, Fangyun Wei, Hongyang Zhang, Yuhui Li","submitted_at":"2024-06-24T17:59:11Z","abstract_excerpt":"Inference with modern Large Language Models (LLMs) is expensive and time-consuming, and speculative sampling has proven to be an effective solution. Most speculative sampling methods such as EAGLE use a static draft tree, implicitly assuming that the acceptance rate of draft tokens depends only on their position. Interestingly, we found that the acceptance rate of draft tokens is also context-dependent. In this paper, building upon EAGLE, we propose EAGLE-2, which introduces a new technique of context-aware dynamic draft tree into drafting modeling. This improvement leverages the fact that the"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.16858","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/2406.16858/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":"2406.16858","created_at":"2026-07-05T08:38:11.847645+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.16858v2","created_at":"2026-07-05T08:38:11.847645+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.16858","created_at":"2026-07-05T08:38:11.847645+00:00"},{"alias_kind":"pith_short_12","alias_value":"KXX4LGLP52GM","created_at":"2026-07-05T08:38:11.847645+00:00"},{"alias_kind":"pith_short_16","alias_value":"KXX4LGLP52GMCPZ7","created_at":"2026-07-05T08:38:11.847645+00:00"},{"alias_kind":"pith_short_8","alias_value":"KXX4LGLP","created_at":"2026-07-05T08:38:11.847645+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":22,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08642","citing_title":"DominoTree: Conditional Tree-Structured Drafting with Domino for Speculative Decoding","ref_index":17,"is_internal_anchor":true},{"citing_arxiv_id":"2606.18394","citing_title":"JetSpec: Breaking the Scaling Ceiling of Speculative Decoding with Parallel Tree Drafting","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12303","citing_title":"From 2D Grids to 1D Tokens: Reforming Shared Representations for Multimodal Image Fusion","ref_index":72,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10935","citing_title":"CLP: Collocation-Length Prediction for Zero-Loss Adaptive Multi-Token Inference","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2606.27550","citing_title":"EntMTP: Accelerating LLM Inference with Entropy Guided Multi Token Prediction","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31315","citing_title":"BlockPilot: Instance-Adaptive Policy Learning for Diffusion-based Speculative Decoding","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00535","citing_title":"DREAM-S: Speculative Decoding with Searchable Drafting and Target-Aware Refinement for Multimodal Generation","ref_index":75,"is_internal_anchor":false},{"citing_arxiv_id":"2501.19201","citing_title":"Efficient Reasoning with Hidden Thinking","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17613","citing_title":"VeriCache: Turning Lossy KV Cache into Lossless LLM Inference","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2509.24328","citing_title":"Speculative Verification: Exploiting Information Gain to Refine Speculative Decoding","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2510.15545","citing_title":"TokenTiming: A Dynamic Alignment Method for Universal Speculative Decoding Model Pairs","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2511.14617","citing_title":"Seer: Online Context Learning for Fast Synchronous LLM Reinforcement Learning","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2601.05524","citing_title":"Double: Breaking the Acceleration Limit via Double Retrieval Speculative Parallelism","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2602.06932","citing_title":"When RL Meets Adaptive Speculative Training: A Unified Training-Serving System","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2603.01581","citing_title":"KERV: Kinematic-Rectified Speculative Decoding for Embodied VLA Models","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2603.17573","citing_title":"HeiSD: Hybrid Speculative Decoding for Embodied Vision-Language-Action Models with Kinematic Awareness","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2603.28049","citing_title":"Drift-AR: Single-Step Visual Autoregressive Generation via Anti-Symmetric Drifting","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03231","citing_title":"CoME-VL: Scaling Complementary Multi-Encoder Vision-Language Learning","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26469","citing_title":"An Empirical Study of Speculative Decoding on Software Engineering Tasks","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10453","citing_title":"SlimSpec: Low-Rank Draft LM-Head for Accelerated Speculative Decoding","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07230","citing_title":"CASCADE: Context-Aware Relaxation for Speculative Image Decoding","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18471","citing_title":"NI Sampling: Accelerating Discrete Diffusion Sampling by Token Order Optimization","ref_index":31,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KXX4LGLP52GMCPZ7SIK7PD35RD","json":"https://pith.science/pith/KXX4LGLP52GMCPZ7SIK7PD35RD.json","graph_json":"https://pith.science/api/pith-number/KXX4LGLP52GMCPZ7SIK7PD35RD/graph.json","events_json":"https://pith.science/api/pith-number/KXX4LGLP52GMCPZ7SIK7PD35RD/events.json","paper":"https://pith.science/paper/KXX4LGLP"},"agent_actions":{"view_html":"https://pith.science/pith/KXX4LGLP52GMCPZ7SIK7PD35RD","download_json":"https://pith.science/pith/KXX4LGLP52GMCPZ7SIK7PD35RD.json","view_paper":"https://pith.science/paper/KXX4LGLP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.16858&json=true","fetch_graph":"https://pith.science/api/pith-number/KXX4LGLP52GMCPZ7SIK7PD35RD/graph.json","fetch_events":"https://pith.science/api/pith-number/KXX4LGLP52GMCPZ7SIK7PD35RD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KXX4LGLP52GMCPZ7SIK7PD35RD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KXX4LGLP52GMCPZ7SIK7PD35RD/action/storage_attestation","attest_author":"https://pith.science/pith/KXX4LGLP52GMCPZ7SIK7PD35RD/action/author_attestation","sign_citation":"https://pith.science/pith/KXX4LGLP52GMCPZ7SIK7PD35RD/action/citation_signature","submit_replication":"https://pith.science/pith/KXX4LGLP52GMCPZ7SIK7PD35RD/action/replication_record"}},"created_at":"2026-07-05T08:38:11.847645+00:00","updated_at":"2026-07-05T08:38:11.847645+00:00"}