Papers › Efficient Streaming Language Models with Attention Sinks
Efficient Streaming Language Models with Attention Sinks
Guangxuan Xiao, Yuandong Tian, Beidi Chen, Song Han, Mike Lewis
Deploying Large Language Models (LLMs) in streaming applications such as multi-round dialogue, where long interactions are expected, is urgently needed but poses two major challenges. Firstly, during the decoding stage, caching previous tokens' Key and Value states (KV) consumes extensive memory. Secondly, popular LLMs cannot generalize to longer texts than the training sequence length. Window attention, where only the most recent KVs are cached, is a natural approach -- but we show that it fails when the text length surpasses the cache size. We observe an interesting phenomenon, namely attention sink, that keeping the KV of initial tokens will largely recover the performance of window attention. In this paper, we first demonstrate that the emergence of attention sink is due to the strong attention scores towards initial tokens as a "sink" even if they are not semantically important. Based on the above analysis, we introduce StreamingLLM, an efficient framework that enables LLMs trained with a finite length attention window to generalize to infinite sequence lengths without any fine-tuning. We show that StreamingLLM can enable Llama-2, MPT, Falcon, and Pythia to perform stable and efficient language modeling with up to 4 million tokens and more. In addition, we discover that adding a placeholder token as a dedicated attention sink during pre-training can further improve streaming deployment. In streaming settings, StreamingLLM outperforms the sliding window recomputation baseline by up to 22.2x speedup. Code and datasets are provided at https://github.com/mit-han-lab/streaming-llm.
In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.
Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.
For agents, on Syntology's MCP service (how to connect):
get_harvested_code_for_paper(arxiv_id="2309.17453")get_code(code_sha256="…")with a sample'scode_sha256from the list belowget_citation_path(paper_1="2309.17453", paper_2="…")with another paper's arXiv id or titleget_concepts_for_paper(arxiv_id="2309.17453")
Code
Syntology 11 code samples harvested from 5 repositories linked to this paper: 8 ran (of which 4 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 4 where Syntology's instrument failed) · 3 unverified. Where Syntology's instrument failed: 2 where our contract draft was wrong; 2 where the contract check could not run; failures of Syntology's instrument, not of the code.
By repository: official repository: 6 samples from 2 repositories, 6 ran (2 with no instrument failure; 4 where Syntology's instrument failed); community (archive-listed): 5 samples from 3 repositories, 2 ran (2 with no instrument failure; 0 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.
No sample is linked to this paper from 1 of its repositories that Syntology harvested for other papers:
- intel/neural-speed, community (archive-listed): harvested for another paper, listed under the repository
Levels written on 2026-09-22 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed. 1 of the 8 samples that ran has no recorded date.
Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.
A paper named beside a repository with no sample linked to this paper is shown with that paper's own run record, not this paper's: “ran” means executed on a synthesized input, not that the code is correct, and “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Papers are listed in arXiv-id order, at most three per repository.
Code Syntology ran Syntology
11 samples harvested: 8 ran (of which 4 constructed an object rather than computing a result; 4 with no instrument failure: 0 honoured, 0 violated, 4 with no contract checked; 4 where Syntology's instrument failed) · 3 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page has not changed since , the first build that kept a record date for it; when this build read Syntology's graph is in the build record. That date is not when the samples ran.
Licence: 2 of the 11 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.
Harvested from 5 repositories linked to this paper, official or community; the list below is grouped by repository, each group headed by its name and what it is to the paper. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.
Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.
Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.
A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.
c2fd5f6408017e25 · report
fixture: returned H2OKVCache · level recorded 2026-09-22 465929486ee1650b · report
fixture: returned StartRecentKVCache · level recorded 2026-09-22 0a11ee057df336ca · report
fixture: signature model:callable/0/?, tokenizer:MinimalTokenizer/0/?, input_ids:Tensor/2/int, past_key_values:NoneType/0/?, max_gen_len:int/0/int · returned list · level recorded, date not recorded 2564045c0e65006b · report
fixture: signature x:Tensor/3/float, start:int/0/int, end:int/0/int · returned Tensor · level recorded 2026-09-22 19c86edfcb34276c · report
fixture: signature x:Tensor/4/float, start:int/0/int, end:int/0/int · returned Tensor · level recorded 2026-09-22 8884022644d5af82 · report
fixture: signature x:Tensor/5/float, start:int/0/int, end:int/0/int · returned Tensor · level recorded 2026-09-22 a71a6943a9cf9d4b · report
fixture: returned SinkCache · level recorded 2026-09-22 1b72723f50079a16 · report
221a10d8ed175d74 · report
fixture: returned AttentionSinkKVCache · level recorded 2026-09-22 Method as code Syntology
What Syntology's graph records as this paper's claimed method, lifted into code: each assembly names its bricks (functions and classes), each with its own role and its own verification level. An assembly is not a check that the code implements the method correctly or reproduces the paper's results. Levels are each brick's own verification level in Syntology's graph (level 0: no recorded run), listed brick by brick and never combined into one level; “ran” means the brick executed on a synthesized input, not that its output is correct, and the instrument figure counts failures of Syntology's instrument, not of the code. Where a model read the code against the paper's claim, each assembly's label verdict is one model's judgement, kept in this page's JSON twin; it is not calibrated, so no card prints it. Siblings are assemblies lifted from this paper in the same batch for the same method: alternative readings kept side by side; none replaces or supersedes another.
4 assemblies on this paper: 4 harvested, 0 generated, listed in 1 groups (siblings lifted in one batch form one group).
4 sibling assemblies for StreamingLLM (harvested), lifted in the same batch on 2026-09-09: alternative readings kept side by side; none replaces or supersedes another.
harvested Assembly for StreamingLLM (graph method: StreamingLLM). 2 bricks; individually on a fixture: 1 ran (of which 1 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified; the one sample that ran constructed an object rather than computing a result. The assembly has not been run end-to-end.
one of 4 siblings
- SinkCache entry · level 1 · ran ·
a71a6943a9cf9d4b - Cache helper · level 0 · unverified ·
1b72723f50079a16
harvested Assembly for StreamingLLM (graph method: StreamingLLM). 4 bricks; individually on a fixture: 4 ran (of which 1 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified. The assembly has not been run end-to-end.
one of 4 siblings
- StartRecentKVCache entry · level 1 · ran ·
465929486ee1650b - slice1d helper · level 1 · ran · our draft was wrong ·
2564045c0e65006b - slice2d helper · level 1 · ran; the contract check could not run (RAISES) ·
19c86edfcb34276c - slice3d helper · level 1 · ran; the contract check could not run (RAISES) ·
8884022644d5af82
harvested Assembly for StreamingLLM (graph method: StreamingLLM). 4 bricks; individually on a fixture: 4 ran (of which 1 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified. The assembly has not been run end-to-end.
one of 4 siblings
- AttentionSinkKVCache entry · level 1 · ran ·
221a10d8ed175d74 - slice1d helper · level 1 · ran · our draft was wrong ·
2564045c0e65006b - slice2d helper · level 1 · ran; the contract check could not run (RAISES) ·
19c86edfcb34276c - slice3d helper · level 1 · ran; the contract check could not run (RAISES) ·
8884022644d5af82
harvested Assembly for StreamingLLM (graph method: StreamingLLM). 7 bricks; individually on a fixture: 4 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 4 where Syntology's instrument failed) · 3 unverified. The assembly has not been run end-to-end.
one of 4 siblings
- KVCacheRecentGlobal entry · level 0 · unverified ·
87d089c843ae1768 - KVCacheHeadConstant helper · level 0 · unverified ·
ae661478c5529ea2 - KVCache model · level 0 · unverified ·
77d2d58362eb96eb - dequantize_tensor helper · level 1 · ran; the contract check could not run ·
5564321cfa1d2c94 - quantize_tensor helper · level 1 · ran; the contract check could not run ·
8d5153e794aa8b15 - unpack_low_bit_tensor helper · level 1 · ran; the contract check could not run ·
910108db225dd9fe - pack_low_bit_tensor helper · level 1 · ran; the contract check could not run ·
f7e4d6a116aaae11
Each brick ends with its code_sha256, Syntology's identity for that exact code; click one to copy the get_code call that returns it. For agents, on Syntology's MCP service: assembly(arxiv_id="2309.17453") returns these assemblies with each brick's role and wiring (how to connect).
Tasks
Results from the paper archive 2025-07-28
No leaderboard rows for this paper in the archive.
Methods
Introduced by this paper: Attention Sinks
Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections