{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,8]],"date-time":"2026-10-08T06:49:23Z","timestamp":1791442163331,"version":"4.3.3"},"reference-count":40,"publisher":"IEEE","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,6]]},"DOI":"10.23919\/isc.2026.11520481","type":"proceedings-article","created":{"date-parts":[[2026,6,8]],"date-time":"2026-06-08T19:48:54Z","timestamp":1780948134000},"page":"1-14","source":"Crossref","is-referenced-by-count":0,"title":["SnapStream: Efficient Long Sequence Decoding on Dataflow Accelerators"],"prefix":"10.23919","author":[{"given":"Jonathan","family":"Li","sequence":"first","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nasim","family":"Farahini","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Evgenii","family":"Iuliugin","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Magnus","family":"Vesterlund","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Christian","family":"H\u00e4ggstr\u00f6m","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Guangtao","family":"Wang","sequence":"additional","affiliation":[{"name":"Meta Platforms, Inc.,USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shubhangi","family":"Upasani","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ayush","family":"Sachdeva","sequence":"additional","affiliation":[{"name":"Cartesia AI,USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Rui","family":"Li","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Faline","family":"Fu","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Chen","family":"Wu","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ayesha","family":"Siddiqua","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"John","family":"Long","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Tuowen","family":"Zhao","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Matheen","family":"Mussadiq","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"H\u00e5kan","family":"Zeffer","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yun","family":"Du","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mingran","family":"Wang","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Qinghua","family":"Li","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Bo","family":"Li","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Urmish","family":"Thakker","sequence":"additional","affiliation":[{"name":"Microsoft AI,USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Raghu","family":"Prabhakar","sequence":"additional","affiliation":[{"name":"SambaNova Systems, Inc.,India"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","volume-title":"Efficient streaming language models with attention sinks","author":"Xiao","year":"2024"},{"key":"ref2","volume-title":"Snapkv: Llm knows what you are looking for before generation","author":"Li","year":"2024"},{"key":"ref3","doi-asserted-by":"crossref","DOI":"10.1145\/3600006.3613165","volume-title":"Efficient memory management for large language model serving with pagedattention","author":"Kwon","year":"2023"},{"key":"ref4","volume-title":"Sglang: Efficient execution of structured language model programs","author":"Zheng","year":"2024"},{"key":"ref5","volume-title":"Streamingllm support?","author":"Chen","year":"2023"},{"key":"ref6","volume-title":"Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning","author":"Guo","year":"2025"},{"key":"ref7","volume-title":"Qwen3 technical report","author":"Yang","year":"2025"},{"key":"ref8","volume-title":"gpt-oss-120b & gpt-oss-20b model card","author":"Agarwal","year":"2025"},{"key":"ref9","volume-title":"Kimi k2: Open agentic intelligence","author":"Kimi","year":"2025"},{"key":"ref10","volume-title":"Gqa: Training generalized multi-query transformer models from multi-head checkpoints","author":"Ainslie","year":"2023"},{"key":"ref11","volume-title":"Deepseek-v2: A strong, economical, and efficient mixture-of-experts language model","author":"Liu","year":"2024"},{"key":"ref12","volume-title":"Gemma 2: Improving open language models at a practical size","author":"Riviere","year":"2024"},{"key":"ref13","volume-title":"H2o: Heavy-hitter oracle for efficient generative inference of large language models","author":"Zhang","year":"2023"},{"key":"ref14","volume-title":"Spargeattention: Accurate and training-free sparse attention accelerating any model inference","author":"Zhang","year":"2025"},{"key":"ref15","volume-title":"Longbench: A bilingual, multitask benchmark for long context understanding","author":"Bai","year":"2024"},{"key":"ref16","volume-title":"Needle in a haystack - pressure testing llms","author":"Kamradt","year":"2023"},{"key":"ref17","volume-title":"Chain-of-thought prompting elicits reasoning in large language models","author":"Wei","year":"2023"},{"key":"ref18","volume-title":"Distserve: Disaggregating prefill and decoding forgoodput-optimized large language model serving","author":"Zhong","year":"2024"},{"key":"ref19","volume-title":"Tensorrt"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1109\/micro61859.2024.00100"},{"key":"ref21","article-title":"Orca: A distributed serving system for transformer-based generative models","volume-title":"USENIX Symposium on Operating Systems Design and Implementation","author":"Yu","year":"2022"},{"key":"ref22","volume-title":"Sarathi: Efficient llm inference by piggybacking decodes with chunked prefills","author":"Agrawal","year":"2023"},{"key":"ref23","doi-asserted-by":"publisher","DOI":"10.1109\/hcs61935.2024.10664717"},{"key":"ref24","volume-title":"The llama 3 herd of models","author":"Grattafiori","year":"2024"},{"key":"ref25","volume-title":"Deepseek-v3 technical report","author":"Liu","year":"2025"},{"key":"ref26","volume-title":"Flashattention: Fast and memory-efficient exact attention with io-awareness","author":"Dao","year":"2022"},{"key":"ref27","volume-title":"Speculative decoding on the sn401 reconfigurable dataflow unit","author":"Nandkar","year":"2025"},{"key":"ref28","volume-title":"Ruler: What\u2019s the real context size of your long-context language models?","author":"Hsieh","year":"2024"},{"key":"ref29","volume-title":"\u221ebench: Extending long context evaluation beyond 100k tokens","author":"Zhang","year":"2024"},{"key":"ref30","volume-title":"Llms know what to drop: Self-attention guided kv cache eviction for efficient long-context inference","author":"Wang","year":"2025"},{"key":"ref31","volume-title":"Attention is all you need","author":"Vaswani","year":"2023"},{"key":"ref32","volume-title":"Generating long sequences with sparse transformers","author":"Child","year":"2019"},{"key":"ref33","volume-title":"Longformer: The long-document transformer","author":"Beltagy","year":"2020"},{"key":"ref34","volume-title":"Reformer: The efficient transformer","author":"Kitaev","year":"2020"},{"key":"ref35","volume-title":"Quest:Query-aware sparsity for efficient long-context llm inference","author":"Tang","year":"2024"},{"key":"ref36","volume-title":"Infllm: Training-free long-context extrapolation for llms with an efficient context memory","author":"Xiao","year":"2024"},{"key":"ref37","volume-title":"Cartridges: Lightweight and general-purpose long context representations via self-study","author":"Eyuboglu","year":"2025"},{"key":"ref38","volume-title":"Titans: Learning to memorize at test time","author":"Behrouz","year":"2024"},{"key":"ref39","volume-title":"Native sparse attention: Hardware-aligned and natively trainable sparse attention","author":"Yuan","year":"2025"},{"key":"ref40","volume-title":"Deepseek-v3.2-exp: Boosting longcontext efficiency with deepseek sparse attention","year":"2025"}],"event":{"name":"ISC High Performance 2026 Research Paper Proceedings (41st International Conference)","location":"Hamburg, Germany","start":{"date-parts":[[2026,6,22]]},"end":{"date-parts":[[2026,6,26]]}},"container-title":["ISC High Performance 2026 Research Paper Proceedings (41st International Conference)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11520448\/11520449\/11520481.pdf?arnumber=11520481","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,8]],"date-time":"2026-10-08T05:50:50Z","timestamp":1791438650000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11520481\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,6]]},"references-count":40,"URL":"https:\/\/doi.org\/10.23919\/isc.2026.11520481","relation":{},"subject":[],"published":{"date-parts":[[2026,6]]}}}