Prefrontal-AI/latent-sparse-attention
Library
Can I trust it?
Worth a closer look before you use it.
Warnings Check
Very new: it was created 7 days ago. Wait a little before installing it.
Stars from real accounts No data
We haven't checked its stars for fakes yet.
Active OK
Last change today.
Licence OK
Apache-2.0: you can use it for free, also in a commercial product. Keep the licence notice.
Quick checks from public data. They help, but they are no guarantee.
What it is
Research code for Latent Sparse Attention, which clusters keys hierarchically in a learned latent space so each query scores cluster centroids and attends only to selected long-range keys plus sink and window tokens. It offers a distilled variant applied to pre-trained models…
Who it's for
Researchers and ML engineers working on subquadratic attention for long-context LLM inference and training
How to use it
Install
uv sync --locked --extra evaluation --extra trainingFrom its README. Check it there before running it.
Ask your AI to set it up
Paste this into Claude, ChatGPT or your coding assistant.
Help me install and set up Prefrontal-AI/latent-sparse-attention in my project. Follow its README: https://github.com/Prefrontal-AI/latent-sparse-attention. Explain each step and check it is safe before running anything.
Star curve
On the radar
Its place among the radar's top 100, day by day, with #1 at the top. Gaps are days it was out of the top 100.
Records
- Best place on RepoAlerts: #66, on 3 Nov 2026
Alternatives
Similar projects named in its profile. Each link searches the directory.
Get alerts for repos like this
We'll email you when something in Models & inference starts rising. Free, once a week, only what matches.
What arrives: one email on Mondays with the repos that match, never more. See last week's. Privacy
For maintainers
Add this badge to your README. It shows the repo's place on the radar, updates every hour and links back to this page.
[](https://repoalerts.com/repo/Prefrontal-AI/latent-sparse-attention?ref=badge)Similar repos
- HazyResearch/ncm
Research code for Next-Concept Modeling, a two-stage framework for parallel text generation. A multiscale residual VQ tokenizer maps sequences to hierarchical discrete codes, and a transformer predict
- token-gate-llms/TokenGate
A research codebase for TokenGate, a drop-in neural component that extends existing trained language models with learned token selection. A gate scores context tokens and keeps the top-K, so only sele
- Interactive-Active-AI/NextTransformer
A research framework and benchmark for comparing Transformer architectures under a fixed training budget. It trains recurrent-depth, sparse and linear attention, residual-stream variants and hybrids o
- apple-aiml-research/ml-rsi-inference-bench
A long-horizon benchmark for coding agents that improve the inference stack of large language models in vLLM or SGLang. Tasks cover supporting a model on day 0, optimizing it after launch, and repairi
- togethercomputer/HybridAgent
Research code for HybridAgent, an LLM agent that alternates between latent reasoning and text-based actions in one generation. It maps last-layer hidden states back to input embeddings via HybridLink,
- SauersML/neural-pairing
Research code that links two language models from different families, Qwen3-4B and AFM-4.5B, through their hidden states. After selected layers each model attends over the other's hidden states and ad