Oct 7, 2026 · Artificial Intelligence

QuantWM: Training-Free 2-bit KV Cache Quantization for Stable Video World Models

Researchers from HIT Shenzhen and NUS propose QuantWM, a training-free 2-bit KV Cache quantization framework that reduces temporal flickering in video world models by protecting attention mechanisms, achieving up to 6.20x compression while improving visual quality over existing methods.

Attention MechanismKV Cache QuantizationLow-Bit Quantization

0 likes · 12 min read

QuantWM: Training-Free 2-bit KV Cache Quantization for Stable Video World Models