QuantWM: Temporally Consistent 2-Bit KV Cache Quantization for World Models and Video Generation

T1Covered by 1 source · first reported by arXiv at 24 Sept, 04:00 UTC

arXiv:2609.26425v2 Announce Type: cross Abstract: KV cache memory has become a major deployment bottleneck for video generation and world models, which motivates low-bit quantization study for efficiency. Existing 2-bit KV cache quantization methods can achieve nearly lossless performance on video…

science

Sources

• T1QuantWM: Temporally Consistent 2-Bit KV Cache Quantization for World Models and Video GenerationarXiv24 Sept, 04:00