Addressable Memory for Video World Models

Hugging Face Daily Papers Papers
video-world-models memory-management key-value-cache long-horizon-generation benchmark compression

Summary

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.

We study visual persistence in interactive video world models. These models rely on a Key-Value (KV) cache as a growing visual memory to carry forward previously generated frames. However, we find that models can no longer reliably address stored content once rollouts extend beyond the training horizon, because temporal Rotary Positional Embeddings (RoPE) offsets then fall outside the range seen during training and the model struggles to retrieve the relevant visual information through attention. Moreover, naively compressing the cache in the RoPE-rotated space corrupts memory by averaging together incompatible positional phases. To address this, we propose WorldTrace, a training-free memory framework for long-horizon visual persistence. WorldTrace keeps compressed memory addressable by assigning each summary slot a distinct, in-distribution virtual position. Within this addressable cache, we study two memory compression approaches: WorldTrace-Field compresses history for temporal coherence, while WorldTrace-Landmark stores verbatim scene traces at detected transitions for episodic recall. We further introduce LoopBench, a benchmark evaluating whether a compressed cache can reconstruct a previously visited scene after a long detour. WorldTrace-Field improves temporal consistency by +15.5%, and WorldTrace-Landmark improves episodic recall by +19.5% on LoopBench, extending visually persistent generation without retraining.
Original Article
View Cached Full Text

Cached at: 08/10/26, 02:12 AM

Paper page - Addressable Memory for Video World Models

Source: https://huggingface.co/papers/2608.07408

Abstract

Westudyvisualpersistenceininteractivevideoworldmodels.ThesemodelsrelyonaKey-Value(KV)cacheasagrowingvisualmemorytocarryforwardpreviouslygeneratedframes.However,wefindthatmodelscannolongerreliablyaddressstoredcontentoncerolloutsextendbeyondthetraininghorizon,becausetemporalRotaryPositionalEmbeddings(RoPE)offsetsthenfalloutsidetherangeseenduringtrainingandthemodelstrugglestoretrievetherelevantvisualinformationthroughattention.Moreover,naivelycompressingthecacheintheRoPE-rotatedspacecorruptsmemorybyaveragingtogetherincompatiblepositionalphases.Toaddressthis,weproposeWorldTrace,atraining-freememoryframeworkforlong-horizonvisualpersistence.WorldTracekeepscompressedmemoryaddressablebyassigningeachsummaryslotadistinct,in-distributionvirtualposition.Withinthisaddressablecache,westudytwomemorycompressionapproaches:WorldTrace-Fieldcompresseshistoryfortemporalcoherence,whileWorldTrace-Landmarkstoresverbatimscenetracesatdetectedtransitionsforepisodicrecall.WefurtherintroduceLoopBench,abenchmarkevaluatingwhetheracompressedcachecanreconstructapreviouslyvisitedsceneafteralongdetour.WorldTrace-Fieldimprovestemporalconsistencyby+15.5%,andWorldTrace-Landmarkimprovesepisodicrecallby+19.5%onLoopBench,extendingvisuallypersistentgenerationwithoutretraining.

View arXiv pageView PDFProject pageAdd to collection

Get this paper in your agent:

hf papers read 2608\.07408

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/2608.07408 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.07408 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.07408 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.