LOCI: Spatial Linear Memory for Streaming World Models

Hugging Face Daily Papers Papers

Summary

LOCI introduces a hybrid spatial memory architecture for streaming video world models, combining key-value caches with a recurrent linear-attention memory conditioned on camera geometry. It reproduces revisited content more faithfully than full-softmax baselines while reducing peak memory by about 30% and enabling constant-memory streaming of long videos.

When a camera revisits a previously observed region, a video world model should reproduce what was there before. This requires both remembering past observations and retrieving the right one for the current viewpoint. Key-value caches preserve visual detail but grow with video length; recurrent memory is compact but compresses history into a fixed-size state, so individual past observations are no longer directly accessible. We introduce LOCI, a hybrid spatial-memory architecture that keeps both representations. In half of the transformer blocks, main attention keeps a key-value cache of past observations; in the other half, it is restricted to the current chunk and complemented by a recurrent linear-attention memory whose reads and writes are conditioned on projective camera geometry, so viewpoint enters both memory addressing and stored content. Recurrent readouts flow into subsequent cache-backed blocks and supply their queries with accumulated scene context. On the public MIND memory benchmark and on held-out recorded trajectories, LOCI reproduces revisited content more faithfully than representative world models and a same-recipe full-softmax model; with full history, it lowers peak memory at equal length by about 30% relative to full softmax. With a bounded bank of retained observations, it streams long videos at constant memory and remains more faithful than full softmax under the same budget.
Original Article
View Cached Full Text

Cached at: 10/02/26, 12:28 PM

Paper page - LOCI: Spatial Linear Memory for Streaming World Models

Source: https://huggingface.co/papers/2609.40222 Published on Sep 30

·

Submitted byhttps://huggingface.co/sum0214

xiajion Oct 2

Abstract

Whenacamerarevisitsapreviouslyobservedregion,avideoworldmodelshouldreproducewhatwastherebefore.Thisrequiresbothrememberingpastobservationsandretrievingtherightoneforthecurrentviewpoint.Key-valuecachespreservevisualdetailbutgrowwithvideolength;recurrentmemoryiscompactbutcompresseshistoryintoafixed-sizestate,soindividualpastobservationsarenolongerdirectlyaccessible.WeintroduceLOCI,ahybridspatial-memoryarchitecturethatkeepsbothrepresentations.Inhalfofthetransformerblocks,mainattentionkeepsakey-valuecacheofpastobservations;intheotherhalf,itisrestrictedtothecurrentchunkandcomplementedbyarecurrentlinear-attentionmemorywhosereadsandwritesareconditionedonprojectivecamerageometry,soviewpointentersbothmemoryaddressingandstoredcontent.Recurrentreadoutsflowintosubsequentcache-backedblocksandsupplytheirquerieswithaccumulatedscenecontext.OnthepublicMINDmemorybenchmarkandonheld-outrecordedtrajectories,LOCIreproducesrevisitedcontentmorefaithfullythanrepresentativeworldmodelsandasame-recipefull-softmaxmodel;withfullhistory,itlowerspeakmemoryatequallengthbyabout30%relativetofullsoftmax.Withaboundedbankofretainedobservations,itstreamslongvideosatconstantmemoryandremainsmorefaithfulthanfullsoftmaxunderthesamebudget.

View arXiv pageView PDFProject pageGitHub2Add to collection

Get this paper in your agent:

hf papers read 2609\.40222

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.40222 in a model README.md to link it from this page.

Datasets citing this paper1

#### sum0214/LOCI-revisit-data Viewer• Updatedabout 2 hours ago • 138 • 42

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.40222 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Latent Spatial Memory for Video World Models

Hugging Face Daily Papers

This paper introduces latent spatial memory for video world models, storing 3D scene information directly in diffusion latent space to avoid costly pixel-space reconstruction. The proposed Mirage framework achieves up to 10.57x faster generation and 55x memory reduction while achieving state-of-the-art performance on WorldScore and RealEstate10K.

Addressable Memory for Video World Models

Hugging Face Daily Papers

This paper introduces WorldTrace, a training-free memory framework for long-horizon video world models that keeps compressed cache addressable, plus LoopBench, a benchmark for episodic recall after long detours. It improves temporal consistency by +15.5% and episodic recall by +19.5% on LoopBench.

Memory as Weights: Internalizing Long-Term History for Streaming Videos

Hugging Face Daily Papers

这篇论文提出 MaLoW(Memory as LoRA Weights),将流式视频的长期历史增量编码进固定大小的 LoRA 权重中,冻结主干模型并把当前观测保留在输入里,从而分离长期记忆与实时感知。该方法在 OVO-Bench 和 StreamingBench 上取得领先表现,并在机器人任务上也提升了成功率。

Memory in Video World Models (6 minute read)

TLDR AI

A best-paper research from NVIDIA and collaborators introduces WorldTrace, a training-free framework that keeps compressed memory addressable in autoregressive video world models by assigning fixed slot-rank positions, enabling coherent long rollouts and long-range recall beyond the training horizon.