# C3ache: Accelerating World Action Models with Cross Inference Chunk Cache

arXiv: https://arxiv.org/abs/2606.08962 · HTML: https://mlsanity.com/paper/2606.08962/

## Key facts

- Cited by 8 papers in the map (data through 14 Oct 2026).
- Placed in Embodied and robot world models, one of 11 directions in World models.
- Builds on 1 of its 34 references that are in the map.
- First posted on arXiv 8 Jun 2026.

## Placed in

- World models / [Embodied and robot world models](https://mlsanity.com/wm/embodied-and-robot-world-models/)
- Vision-language-action models / [Efficient VLA inference and adaptation](https://mlsanity.com/vla/efficient-vla-inference-and-adaptation/)

## Builds on (most cited first, 1 of 1)

- [Wan: Open and Advanced Large-Scale Video Generative Models](https://mlsanity.com/paper/2503.20314/) (2025)

## Built on by (most cited first, 8 of 8)

- [Test-Time Scaling for World Action Models via Zero-Shot Geometric Evaluation](https://mlsanity.com/paper/2607.17454/) (2026)
- [PhyAI: Real-Time Physical AI at the Edge, Scalable Rollouts in the Cloud](https://mlsanity.com/paper/2608.03682/) (2026)
- [ActionSplice: In-Flight Action Editing for Interactive World Models](https://mlsanity.com/paper/2609.08230/) (2026)
- [Efficient World Action Model Inference with Adaptive Intermediate States](https://mlsanity.com/paper/2609.34608/) (2026)
- [Sparse-WAM: Accelerating World Action Models via Action-Guided Sparse Imagination](https://mlsanity.com/paper/2609.38984/) (2026)
- [WAMJET: A Harness for World Action Model Acceleration](https://mlsanity.com/paper/2610.03797/) (2026)
- [WAM-Cache: Staleness-Bounded KV Reuse for Efficient World Action Models](https://mlsanity.com/paper/2610.11401/) (2026)
- [MARS: Systematic Multi-Level Acceleration for World Action Models](https://mlsanity.com/paper/2610.17141/) (2026)

## Abstract (arXiv)

World Action Models (WAMs) generalize better than standard Vision-Language-Action (VLA) policies to novel motions and environments, because a video-modeling objective lets them learn from abundant unlabeled video rather than scarce labeled robot demonstrations. This generalization is computationally expensive. To complete a task, a WAM runs over multiple inference chunks, and each chunk requires a costly denoising process. Existing acceleration methods reduce this cost by caching and reusing computation within a single chunk's denoising trajectory. Our empirical analysis reveals a substantial source of redundancy they overlook: redundancy across chunks. When a robot executes a smooth behavior, the residuals computed at a given denoising step are strongly correlated from one chunk to the next. We introduce C3ache, a training-free method that caches and reuses these residuals across inference chunks at the same denoising step. Experiments on benchmarks with a Fast-WAM backbone show that C3ache achieves up to a 2.5× speedup in total wall-clock inference time, with negligible degradation in task success rate.