{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,11,11]],"date-time":"2025-11-11T06:03:44Z","timestamp":1762841024923,"version":"build-2065373602"},"reference-count":12,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"5","license":[{"start":{"date-parts":[[2025,9,1]],"date-time":"2025-09-01T00:00:00Z","timestamp":1756684800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2025,9,1]],"date-time":"2025-09-01T00:00:00Z","timestamp":1756684800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,9,1]],"date-time":"2025-09-01T00:00:00Z","timestamp":1756684800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Micro"],"published-print":{"date-parts":[[2025,9]]},"DOI":"10.1109\/mm.2025.3592570","type":"journal-article","created":{"date-parts":[[2025,7,29]],"date-time":"2025-07-29T18:30:16Z","timestamp":1753813816000},"page":"56-66","source":"Crossref","is-referenced-by-count":0,"title":["Speculative Decoding on the SN40L Reconfigurable Dataflow Unit"],"prefix":"10.1109","volume":"45","author":[{"ORCID":"https:\/\/orcid.org\/0009-0002-9442-4228","authenticated-orcid":false,"given":"Darshan","family":"Gandhi","sequence":"first","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]},{"ORCID":"https:\/\/orcid.org\/0009-0004-5810-2162","authenticated-orcid":false,"given":"Pushkar","family":"Nandkar","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]},{"ORCID":"https:\/\/orcid.org\/0009-0000-5072-4152","authenticated-orcid":false,"given":"Nasim","family":"Farahini","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]},{"ORCID":"https:\/\/orcid.org\/0009-0001-3947-0478","authenticated-orcid":false,"given":"H\u00e5kan","family":"Zeffer","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7406-1410","authenticated-orcid":false,"given":"John","family":"Long","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]},{"ORCID":"https:\/\/orcid.org\/0009-0000-6582-9416","authenticated-orcid":false,"given":"Samuel","family":"Rydh","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]},{"ORCID":"https:\/\/orcid.org\/0009-0003-9118-385X","authenticated-orcid":false,"given":"Matheen","family":"Musaddiq","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-6347-7135","authenticated-orcid":false,"given":"Tuowen","family":"Zhao","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]},{"ORCID":"https:\/\/orcid.org\/0009-0007-6959-765X","authenticated-orcid":false,"given":"Joshua","family":"Brot","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]},{"given":"Reid","family":"Goodbar","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]},{"ORCID":"https:\/\/orcid.org\/0009-0001-6723-4384","authenticated-orcid":false,"given":"Yun","family":"Du","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]},{"ORCID":"https:\/\/orcid.org\/0009-0006-8119-2201","authenticated-orcid":false,"given":"Mingran","family":"Wang","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0230-4377","authenticated-orcid":false,"given":"Raghu","family":"Prabhakar","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc., Palo Alto, CA, USA"}]}],"member":"263","reference":[{"article-title":"The Llama 3 herd of models","year":"2024","author":"Grattafiori","key":"ref1"},{"volume-title":"Llama4 specifications","year":"2025","key":"ref2"},{"year":"2024","key":"ref3","article-title":"DeepSeek-V3 technical report"},{"volume-title":"Scaling test-time compute with open models","year":"2025","author":"Beeching","key":"ref4"},{"year":"2025","key":"ref5","article-title":"DeepSeek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning"},{"volume-title":"Prompt caching","year":"2025","key":"ref6"},{"article-title":"Accelerating large language model decoding with speculative sampling","year":"2023","author":"Chen","key":"ref7"},{"volume-title":"NVIDIA DGX H100 datasheet","year":"2025","key":"ref8"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1109\/micro61859.2024.00100"},{"volume-title":"Independent analysis of AI models and API providers","year":"2025","key":"ref10"},{"volume-title":"Datasets","year":"2025","key":"ref11"},{"volume-title":"NVIDIA: TensorRT-LLM","key":"ref12"}],"container-title":["IEEE Micro"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/40\/11236092\/11098858.pdf?arnumber=11098858","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,11,11]],"date-time":"2025-11-11T06:00:51Z","timestamp":1762840851000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11098858\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,9]]},"references-count":12,"journal-issue":{"issue":"5"},"URL":"https:\/\/doi.org\/10.1109\/mm.2025.3592570","relation":{},"ISSN":["0272-1732","1937-4143"],"issn-type":[{"type":"print","value":"0272-1732"},{"type":"electronic","value":"1937-4143"}],"subject":[],"published":{"date-parts":[[2025,9]]}}}