WAM-Cache: Staleness-Bounded KV Reuse for Efficient World Action Models

Key points

• 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.1
• 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.1
• 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.1
• 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.1

Sources (1)

• [1]WAM-Cache: Staleness-Bounded KV Reuse for Efficient World Action Models
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 07:33 AM

  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.

Extractive summary: sentences quoted from the sources.