Complementary Future Token Pruning for World Action Models
We propose CF-Prune, a training-free framework for pruning future-frame tokens in pretrained WAMs. Our key insight is that future visual predictions support action generation and therefore need not preserve every visual detail required for high-fidelity video synthesis.
ProofPaper ↗
Key points
- World action models (WAMs) jointly predict future visual states and robot actions, but processing dense future-frame tokens incurs substantial inference latency.
- CF-Prune combines instruction and action attention to identify task-relevant tokens during the initial denoising forward pass, even when future-frame and action representations remain highly noisy.
- An entropy-based phase identification mechanism further retains dense computation during precision phases to mitigate pruning-induced performance degradation.
- Experiments on RoboLab, RoboCasa, and real-world manipulation tasks demonstrate consistent inference speedups.
Sources (1)
- [1]Complementary Future Token Pruning for World Action ModelsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 14, 10:39 AM
We propose CF-Prune, a training-free framework for pruning future-frame tokens in pretrained WAMs. Our key insight is that future visual predictions support action generation and therefore need not preserve every visual detail required for high-fidelity video synthesis.
World action models (WAMs) jointly predict future visual states and robot actions, but processing dense future-frame tokens incurs substantial inference latency.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 13, 2026Run fast, private agents on-device with LFM2.5-24B-A2B
- Oct 13, 2026Gemini 4 Argon for Robotics
- Oct 13, 2026inclusionAI/Ling-3.1-flash
- Oct 13, 2026Hybrid Reasoning Agents
- Oct 13, 2026Fish Audio Drama 3 Technical Report
- Oct 13, 2026CLASHarness: Benchmarking Multi-Agent Conflict in Computer Use via a Controllable and Verifiable Generation Pipeline