{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:4PX7HNECFUZR42EV6XT7VVNC73","short_pith_number":"pith:4PX7HNEC","canonical_record":{"source":{"id":"2508.08192","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-08-11T17:11:26Z","cross_cats_sorted":[],"title_canon_sha256":"62a25f22164ff93cfc5149b1bf77445691a59e5ce44657201919a81d9d1147af","abstract_canon_sha256":"5eb03132a9e98f0aa381e19cf7a843146223c4a4212b8aef428165e97110e7a8"},"schema_version":"1.0"},"canonical_sha256":"e3eff3b4822d331e6895f5e7fad5a2fee6732bf682130f9788d7996d2a5c09c7","source":{"kind":"arxiv","id":"2508.08192","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2508.08192","created_at":"2026-07-05T11:52:09Z"},{"alias_kind":"arxiv_version","alias_value":"2508.08192v1","created_at":"2026-07-05T11:52:09Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.08192","created_at":"2026-07-05T11:52:09Z"},{"alias_kind":"pith_short_12","alias_value":"4PX7HNECFUZR","created_at":"2026-07-05T11:52:09Z"},{"alias_kind":"pith_short_16","alias_value":"4PX7HNECFUZR42EV","created_at":"2026-07-05T11:52:09Z"},{"alias_kind":"pith_short_8","alias_value":"4PX7HNEC","created_at":"2026-07-05T11:52:09Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:4PX7HNECFUZR42EV6XT7VVNC73","target":"record","payload":{"canonical_record":{"source":{"id":"2508.08192","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-08-11T17:11:26Z","cross_cats_sorted":[],"title_canon_sha256":"62a25f22164ff93cfc5149b1bf77445691a59e5ce44657201919a81d9d1147af","abstract_canon_sha256":"5eb03132a9e98f0aa381e19cf7a843146223c4a4212b8aef428165e97110e7a8"},"schema_version":"1.0"},"canonical_sha256":"e3eff3b4822d331e6895f5e7fad5a2fee6732bf682130f9788d7996d2a5c09c7","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:52:09.661564Z","signature_b64":"gkDcTj424fDkxuhHs52GyN/nZL7EsUEzEL36Vs9BEr4q4NfzmEKhP20mrNILWQuqaxjG2PEjW6UdmfgT2sYuBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e3eff3b4822d331e6895f5e7fad5a2fee6732bf682130f9788d7996d2a5c09c7","last_reissued_at":"2026-07-05T11:52:09.661060Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:52:09.661060Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2508.08192","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:52:09Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"WIMGqbqXxl5PqYj685IUGa9HxL688Xy4rnGbESs+7HvgAx2zRHYqOGsKadAmQ+rURdjYn1UlTHJJHXxGihr3BA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T06:51:59.321332Z"},"content_sha256":"f6ee8dce6309b756df1e1faddb576806b3cc24b3f7cfad99ef42c3aa634e006b","schema_version":"1.0","event_id":"sha256:f6ee8dce6309b756df1e1faddb576806b3cc24b3f7cfad99ef42c3aa634e006b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:4PX7HNECFUZR42EV6XT7VVNC73","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Efficient Speculative Decoding for Llama at Scale: Challenges and Solutions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Adolfo Victoria, Aya Ibrahim, Bangsheng Tang, Bram Wasti, Carl Chengyan Fu, Changkyu Kim, Daniel Haziza, Emily Guo, Fei Kou, Fei Sun, Giancarlo Delfin, Grigory Sizov, Haoci Zhang, Jaewon Lee, Jason Park, Jeremy Reizenstein, Jialin Ouyang, Jianyu Huang, Jiawen Liu, Jie You, Lu Fang, Qirui Yang, Quinn Zhu, Ria Verma, Sachin Mehta, Shengyong Cai, Vlad Mihailescu, Wei Wei, Xiaodong Wang, Xingwen Guo, Xingyu Liu, Yan Cui, Yanjun Zhou, Ye Hu, Yejin Lee, Yunlu Li, Zhiwei Zhao, Zixi Qi","submitted_at":"2025-08-11T17:11:26Z","abstract_excerpt":"Speculative decoding is a standard method for accelerating the inference speed of large language models. However, scaling it for production environments poses several engineering challenges, including efficiently implementing different operations (e.g., tree attention and multi-round speculative decoding) on GPU. In this paper, we detail the training and inference optimization techniques that we have implemented to enable EAGLE-based speculative decoding at a production scale for Llama models. With these changes, we achieve a new state-of-the-art inference latency for Llama models. For example"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.08192","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/2508.08192/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T11:52:09Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"EHuXUh6600m79ucUA259YhRkvkF/L4rEQXEqdpLIvsyUmsokDgTc98LrszINtaBkJ+fai4R45r69k0XWqu7MBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-06T06:51:59.321856Z"},"content_sha256":"06e3dca2ed9b2a792c0d1ae36a3d123577ca19407b14ea2472580d715ebd670f","schema_version":"1.0","event_id":"sha256:06e3dca2ed9b2a792c0d1ae36a3d123577ca19407b14ea2472580d715ebd670f"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/4PX7HNECFUZR42EV6XT7VVNC73/bundle.json","state_url":"https://pith.science/pith/4PX7HNECFUZR42EV6XT7VVNC73/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/4PX7HNECFUZR42EV6XT7VVNC73/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-06T06:51:59Z","links":{"resolver":"https://pith.science/pith/4PX7HNECFUZR42EV6XT7VVNC73","bundle":"https://pith.science/pith/4PX7HNECFUZR42EV6XT7VVNC73/bundle.json","state":"https://pith.science/pith/4PX7HNECFUZR42EV6XT7VVNC73/state.json","well_known_bundle":"https://pith.science/.well-known/pith/4PX7HNECFUZR42EV6XT7VVNC73/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:4PX7HNECFUZR42EV6XT7VVNC73","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"5eb03132a9e98f0aa381e19cf7a843146223c4a4212b8aef428165e97110e7a8","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-08-11T17:11:26Z","title_canon_sha256":"62a25f22164ff93cfc5149b1bf77445691a59e5ce44657201919a81d9d1147af"},"schema_version":"1.0","source":{"id":"2508.08192","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2508.08192","created_at":"2026-07-05T11:52:09Z"},{"alias_kind":"arxiv_version","alias_value":"2508.08192v1","created_at":"2026-07-05T11:52:09Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.08192","created_at":"2026-07-05T11:52:09Z"},{"alias_kind":"pith_short_12","alias_value":"4PX7HNECFUZR","created_at":"2026-07-05T11:52:09Z"},{"alias_kind":"pith_short_16","alias_value":"4PX7HNECFUZR42EV","created_at":"2026-07-05T11:52:09Z"},{"alias_kind":"pith_short_8","alias_value":"4PX7HNEC","created_at":"2026-07-05T11:52:09Z"}],"graph_snapshots":[{"event_id":"sha256:06e3dca2ed9b2a792c0d1ae36a3d123577ca19407b14ea2472580d715ebd670f","target":"graph","created_at":"2026-07-05T11:52:09Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2508.08192/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Speculative decoding is a standard method for accelerating the inference speed of large language models. However, scaling it for production environments poses several engineering challenges, including efficiently implementing different operations (e.g., tree attention and multi-round speculative decoding) on GPU. In this paper, we detail the training and inference optimization techniques that we have implemented to enable EAGLE-based speculative decoding at a production scale for Llama models. With these changes, we achieve a new state-of-the-art inference latency for Llama models. For example","authors_text":"Adolfo Victoria, Aya Ibrahim, Bangsheng Tang, Bram Wasti, Carl Chengyan Fu, Changkyu Kim, Daniel Haziza, Emily Guo, Fei Kou, Fei Sun, Giancarlo Delfin, Grigory Sizov, Haoci Zhang, Jaewon Lee, Jason Park, Jeremy Reizenstein, Jialin Ouyang, Jianyu Huang, Jiawen Liu, Jie You, Lu Fang, Qirui Yang, Quinn Zhu, Ria Verma, Sachin Mehta, Shengyong Cai, Vlad Mihailescu, Wei Wei, Xiaodong Wang, Xingwen Guo, Xingyu Liu, Yan Cui, Yanjun Zhou, Ye Hu, Yejin Lee, Yunlu Li, Zhiwei Zhao, Zixi Qi","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-08-11T17:11:26Z","title":"Efficient Speculative Decoding for Llama at Scale: Challenges and Solutions"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.08192","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:f6ee8dce6309b756df1e1faddb576806b3cc24b3f7cfad99ef42c3aa634e006b","target":"record","created_at":"2026-07-05T11:52:09Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"5eb03132a9e98f0aa381e19cf7a843146223c4a4212b8aef428165e97110e7a8","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-08-11T17:11:26Z","title_canon_sha256":"62a25f22164ff93cfc5149b1bf77445691a59e5ce44657201919a81d9d1147af"},"schema_version":"1.0","source":{"id":"2508.08192","kind":"arxiv","version":1}},"canonical_sha256":"e3eff3b4822d331e6895f5e7fad5a2fee6732bf682130f9788d7996d2a5c09c7","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"e3eff3b4822d331e6895f5e7fad5a2fee6732bf682130f9788d7996d2a5c09c7","first_computed_at":"2026-07-05T11:52:09.661060Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:52:09.661060Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"gkDcTj424fDkxuhHs52GyN/nZL7EsUEzEL36Vs9BEr4q4NfzmEKhP20mrNILWQuqaxjG2PEjW6UdmfgT2sYuBg==","signature_status":"signed_v1","signed_at":"2026-07-05T11:52:09.661564Z","signed_message":"canonical_sha256_bytes"},"source_id":"2508.08192","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f6ee8dce6309b756df1e1faddb576806b3cc24b3f7cfad99ef42c3aa634e006b","sha256:06e3dca2ed9b2a792c0d1ae36a3d123577ca19407b14ea2472580d715ebd670f"],"state_sha256":"3ebea07944f72865d8cc5ec283543a8110a1a8c6f3683f0bbfb94fdef400bef1"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ZAsTJLjZoiWYAfMp8PvQHfrZPQ1ugVs10hF1UYO0RxLNlkkH6PL5uXaYJlCPclrSqlEXJdUYKI+IQ82XfxXCCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-06T06:51:59.328106Z","bundle_sha256":"3d84246f6d37124fa125fc82dfcbfaa657bfbe37dc4669e1d53c9ac950918469"}}