Papers › Hungry Hungry Hippos: Towards Language Modeling with State Space Models
Hungry Hungry Hippos: Towards Language Modeling with State Space Models
Daniel Y. Fu, Tri Dao, Khaled K. Saab, Armin W. Thomas, Atri Rudra, Christopher Ré
State space models (SSMs) have demonstrated state-of-the-art sequence modeling performance in some modalities, but underperform attention in language modeling. Moreover, despite scaling nearly linearly in sequence length instead of quadratically, SSMs are still slower than Transformers due to poor hardware utilization. In this paper, we make progress on understanding the expressivity gap between SSMs and attention in language modeling, and on reducing the hardware barrier between SSMs and attention. First, we use synthetic language modeling tasks to understand the gap between SSMs and attention. We find that existing SSMs struggle with two capabilities: recalling earlier tokens in the sequence and comparing tokens across the sequence. To understand the impact on language modeling, we propose a new SSM layer, H3, that is explicitly designed for these abilities. H3 matches attention on the synthetic languages and comes within 0.4 PPL of Transformers on OpenWebText. Furthermore, a hybrid 125M-parameter H3-attention model that retains two attention layers surprisingly outperforms Transformers on OpenWebText by 1.0 PPL. Next, to improve the efficiency of training SSMs on modern hardware, we propose FlashConv. FlashConv uses a fused block FFT algorithm to improve efficiency on sequences up to 8K, and introduces a novel state passing algorithm that exploits the recurrent properties of SSMs to scale to longer sequences. FlashConv yields 2× speedup on the long-range arena benchmark and allows hybrid language models to generate text 2.4× faster than Transformers. Using FlashConv, we scale hybrid H3-attention language models up to 2.7B parameters on the Pile and find promising initial results, achieving lower perplexity than Transformers and outperforming Transformers in zero- and few-shot learning on a majority of tasks in the SuperGLUE benchmark.
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="2212.14052")get_code(code_sha256="…")with a sample'scode_sha256from the list belowget_citation_path(paper_1="2212.14052", paper_2="…")with another paper's arXiv id or titleget_concepts_for_paper(arxiv_id="2212.14052")
Code
Syntology 15 code samples harvested from 2 repositories linked to this paper: 7 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 1 honoured, 0 violated, 2 with no contract checked; 4 where Syntology's instrument failed) · 8 unverified. Where Syntology's instrument failed: 3 where our contract draft was wrong; 1 where the contract check could not run; failures of Syntology's instrument, not of the code.
By repository: official repository: 1 sample from 1 repository, 1 ran (0 with no instrument failure; 1 where Syntology's instrument failed); community (archive-listed): 14 samples from 1 repository, 6 ran (3 with no instrument failure; 3 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:
- lindermanlab/S5, community (archive-listed): harvested for another paper, listed under the repository
Levels written between 2026-09-22 and 2026-09-24 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed. 1 of the 7 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
15 samples harvested: 7 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 1 honoured, 0 violated, 2 with no contract checked; 4 where Syntology's instrument failed) · 8 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: 0 of the 15 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 2 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.
5dd28d20c60f0c19 · report
fixture: returned Tensor · level recorded, date not recorded 497dbc1b6483fce5 · report
fixture: signature L:int/0/int, state:NoneType/0/?, rate:float/0/float · returned tuple · level recorded 2026-09-22 ca63d87a582adba1 · report
fixture: signature state:NoneType/0/?, rate:float/0/float, L:int/0/int · returned tuple · level recorded 2026-09-24 cde6f4e2d1c79e3a · report
fixture: signature measure:str/0/str, N:int/0/int, rank:int/0/int, dtype:dtype/0/? · returned tuple · level recorded 2026-09-22 55a38587423268a5 · report
fixture: signature L:int/0/int, A:Tensor/3/float, v:Tensor/3/float · returned tuple · level recorded 2026-09-22 7f226e8168260082 · report
fixture: signature measure:str/0/str, N:int/0/int, rank:int/0/int, dtype:dtype/0/? · returned Tensor · level recorded 2026-09-22 93675235174e1ee4 · report
fixture: signature measure:str/0/str, N:int/0/int, alpha:float/0/float, beta:float/0/float · returned tuple · level recorded 2026-09-22 60be78d8667c0da7 · report
c73230013cf47ccb · report
871ae87aa2c2ffb6 · report
37bc70e26a2b9b9e · report
2f58a2eaf4ce01c9 · report
3d83038bb3d6c578 · report
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.
1 assembly on this paper: 1 harvested, 0 generated.
harvested Assembly for H3 (graph method: H3). 18 bricks; individually on a fixture: 8 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 1 honoured, 0 violated, 2 with no contract checked; 5 where Syntology's instrument failed) · 10 unverified. The assembly has not been run end-to-end.
- H3 entry · level 0 · unverified ·
b100cfc9148d4a59 - SSKernel helper · level 0 · unverified ·
1d9eb34a580a6143 - mul_sum helper · level 1 · ran; the contract check could not run (RAISES) ·
5dd28d20c60f0c19 - EMAKernel helper · level 1 · ran ·
497dbc1b6483fce5 - SSKernelDiag helper · level 0 · unverified ·
60be78d8667c0da7 - SSKernelShift helper · level 1 · ran ·
ca63d87a582adba1 - combination helper · level 0 · unverified ·
c73230013cf47ccb - dplr helper · level 0 · unverified ·
871ae87aa2c2ffb6 - power helper · level 1 · ran · our draft was wrong ·
55a38587423268a5 - OptimModule helper · level 0 · unverified ·
6f2ce45de2079602 - log_vandermonde helper · level 0 · unverified ·
37bc70e26a2b9b9e - log_vandermonde_transpose helper · level 0 · unverified ·
2f58a2eaf4ce01c9 - ssm helper · level 0 · unverified ·
3d83038bb3d6c578 - nplr helper · level 1 · ran · our draft was wrong ·
cde6f4e2d1c79e3a - _broadcast_dims helper · level 0 · unverified ·
0c3303405c8151e9 - rank_correction helper · level 1 · ran · honoured contract ·
7f226e8168260082 - transition helper · level 1 · ran · our draft was wrong ·
93675235174e1ee4 - get_logger helper · level 1 · ran · our draft was wrong ·
61139ec62260b411
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="2212.14052") returns these assemblies with each brick's role and wiring (how to connect).
Tasks
1 archive task tag without a task page not shown.
Results from the paper archive 2025-07-28
| Task | Dataset | Model | Metric | Value | Rank at snapshot | Leaderboard | Report |
|---|---|---|---|---|---|---|---|
| Coreference Resolution | Winograd Schema Challenge | H3 125M (3-shot, rank classification) | Accuracy | 63.5 | #46 of 82 | Archive leaderboard | report |
| Coreference Resolution | Winograd Schema Challenge | H3 125M (0-shot, rank classification) | Accuracy | 61.5 | #55 of 82 | Archive leaderboard | report |
| Coreference Resolution | Winograd Schema Challenge | Hybrid H3 125M (3-shot, logit scoring) | Accuracy | 43.3 | #78 of 82 | Archive leaderboard | report |
| Language Modelling | The Pile | Hybrid H3 125M | Test perplexity | 10.2 | #29 of 39 | Archive leaderboard | report |
| Language Modelling | The Pile | Transformer 125M | Test perplexity | 10.7 | #31 of 39 | Archive leaderboard | report |
| Language Modelling | WikiText-103 | Hybrid H3 (2.7B) | Number of params | 2700M | #2 of 89 | Archive leaderboard | report |
| Language Modelling | WikiText-103 | Hybrid H3 (2.7B) | Test perplexity | 10.6 | #2 of 89 | Archive leaderboard | report |
| Language Modelling | WikiText-103 | Hybrid H3 (1.3B) | Number of params | 1300M | #6 of 89 | Archive leaderboard | report |
| Language Modelling | WikiText-103 | Hybrid H3 (1.3B) | Test perplexity | 12.5 | #6 of 89 | Archive leaderboard | report |
| Language Modelling | WikiText-103 | Hybrid H3 (355M) | Number of params | 355M | #18 of 89 | Archive leaderboard | report |
| Language Modelling | WikiText-103 | Hybrid H3 (355M) | Test perplexity | 16.9 | #18 of 89 | Archive leaderboard | report |
| Language Modelling | WikiText-103 | Hybrid H3 125M | Test perplexity | 18.5 | #38 of 89 | Archive leaderboard | report |
| Language Modelling | WikiText-103 | Hybrid H3 (125M) | Number of params | 125M | #53 of 89 | Archive leaderboard | report |
| Language Modelling | WikiText-103 | Hybrid H3 (125M) | Test perplexity | 23.7 | #53 of 89 | Archive leaderboard | report |
| Natural Language Inference | RTE | Hybrid H3 125M (0-shot, logit scoring) | Accuracy | 59.2% | #74 of 90 | Archive leaderboard | report |
| Natural Language Inference | RTE | Hybrid H3 125M (3-shot, logit scoring) | Accuracy | 58.1% | #77 of 90 | Archive leaderboard | report |
| Natural Language Inference | RTE | Hybrid H3 125M (3-shot, rank classification) | Accuracy | 58.1% | #78 of 90 | Archive leaderboard | report |
| Natural Language Inference | RTE | H3 125M (0-shot, rank classification) | Accuracy | 53.1% | #88 of 90 | Archive leaderboard | report |
| Natural Language Inference | RTE | H3 125M (3-shot, rank classification) | Accuracy | 52.3% | #90 of 90 | Archive leaderboard | report |
| Question Answering | BoolQ | Hybrid H3 1.3B (0-shot, logit scoring) | Accuracy | 61.7 | #52 of 65 | Archive leaderboard | report |
| Question Answering | BoolQ | Hybrid H3 2.7B (3-shot, logit scoring) | Accuracy | 60.6 | #55 of 65 | Archive leaderboard | report |
| Question Answering | BoolQ | Hybrid H3 125M (0-shot, logit scoring) | Accuracy | 59.6 | #59 of 65 | Archive leaderboard | report |
| Question Answering | BoolQ | Hybrid H3 125M (3-shot, logit scoring) | Accuracy | 56.1 | #61 of 65 | Archive leaderboard | report |
| Question Answering | BoolQ | Hybrid H3 125M (3-shot, rank classification) | Accuracy | 56.1 | #62 of 65 | Archive leaderboard | report |
| Question Answering | COPA | Hybrid H3 2.7B (0-shot, logit scoring) | Accuracy | 81 | #35 of 60 | Archive leaderboard | report |
| Question Answering | COPA | Hybrid H3 2.7B (3-shot, logit scoring) | Accuracy | 77 | #42 of 60 | Archive leaderboard | report |
| Question Answering | COPA | Hybrid H3 125M (0-shot, logit scoring) | Accuracy | 67 | #51 of 60 | Archive leaderboard | report |
| Question Answering | COPA | Hybrid H3 125M (0-shot, rank classification) | Accuracy | 67 | #52 of 60 | Archive leaderboard | report |
| Question Answering | COPA | H3 125M (0-shot, rank classification) | Accuracy | 51 | #59 of 60 | Archive leaderboard | report |
| Question Answering | MultiRC | Hybrid H3 355M (3-shot, logit scoring) | EM | 59.7 | #27 of 30 | Archive leaderboard | report |
| Question Answering | MultiRC | Hybrid H3 355M (0-shot, logit scoring) | EM | 59.5 | #28 of 30 | Archive leaderboard | report |
| Question Answering | MultiRC | Hybrid H3 125M (0-shot, logit scoring) | EM | 51.4 | #29 of 30 | Archive leaderboard | report |
| Question Answering | MultiRC | Hybrid H3 125M (3-shot, logit scoring) | EM | 48.9 | #30 of 30 | Archive leaderboard | report |
| Word Sense Disambiguation | Words in Context | Hybrid H3 125M (0-shot, logit scoring) | Accuracy | 51.4 | #29 of 37 | Archive leaderboard | report |
| Word Sense Disambiguation | Words in Context | Hybrid H3 125M (0-shot, rank classification) | Accuracy | 51.4 | #30 of 37 | Archive leaderboard | report |
| Word Sense Disambiguation | Words in Context | Hybrid H3 125M (3-shot, logit scoring) | Accuracy | 49.1 | #37 of 37 | Archive leaderboard | report |
Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.
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