Metraipolis › archive › dc_53ef07f1e882
WAM-Cache: Staleness-Bounded KV Reuse for Efficient World Action Models — arXiv:2610.11401
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.11401
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-09T06:24:03.260Z
- SHA-256 of the archived text
0d7cf4dfe786295a8ed4b0fd6cfa9c0ff1df93aa79eb19b885b6d9b92b19dea0- Stable URL of this page
https://metraipolis.com/documento/dc_53ef07f1e882
How to cite
WAM-Cache: Staleness-Bounded KV Reuse for Efficient World Action Models — arXiv:2610.11401 — arXiv (en), https://arxiv.org/abs/2610.11401 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_53ef07f1e882 (Metraipolis archive dc_53ef07f1e882, sha256 0d7cf4dfe786295a8ed4b0fd6cfa9c0ff1df93aa79eb19b885b6d9b92b19dea0)
BibTeX
@misc{dc53ef07f1e882,
title = {WAM-Cache: Staleness-Bounded KV Reuse for Efficient World Action Models — arXiv:2610.11401},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.11401},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_53ef07f1e882; sha256 0d7cf4dfe786295a8ed4b0fd6cfa9c0ff1df93aa79eb19b885b6d9b92b19dea0},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.11401. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
World Action Models (WAMs) enable generalist robot manipulation by conditioning an action expert on representations from a pretrained video Diffusion Transformer (DiT). In closed-loop control, the video DiT runs at every chunk to encode the current observation into layerwise key-value (KV) pairs that the action expert queries. This prefill dominates the per-chunk computational cost, yet existing training-free accelerations leave it fully dense. We present WAM-Cache, a training-free framework that retains layerwise key-value representations across chunks and recomputes only a sparse refresh set of tokens. Crucially, we find that the intuitive heuristic of refreshing visually drifted tokens plateaus far below the dense baseline, even with an oracle predicting ground-truth KV drift. Downstream action accuracy is instead governed by where the action expert attends, not by what moved. WAM-Cache therefore selects the refresh set by uniting the action expert's cross-attention with visual latent surprise, complemented by a strict age bound that suppresses compounding error. On Fast-WAM, WAM-Cache cuts video DiT prefill FLOPs by 32-42% across RoboTwin 2.0, LIBERO, and real-world experiments, while staying within 0.7-1.8 percentage points of the dense policy in simulation and 2.5 points on a real robot. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno). Categorie arXiv: Robotics (cs.RO) ; Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.11401 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-09T05:50:32.010Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).