KVCacheis1BitPerChannel: EfficientLarge LanguageModelInferencewithCoupledQuantization

Feb-7-2026, 10:04:28 GMT–Neural Information Processing Systems

Furthermore, we demonstrate that CQ can preservemodel quality reasonably with KV cache quantizeddownto1bit.

large language model, machine learning, natural language, (19 more...)

Neural Information Processing Systems

Feb-7-2026, 10:04:28 GMT

Conferences PDF

Country:
- North America > United States
  - New Jersey > Hudson County
    - Hoboken (0.04)
  - Texas > Harris County
    - Houston (0.04)

Genre:
- Research Report (0.46)

Technology:
- Information Technology > Artificial Intelligence
  - Machine Learning (1.00)
  - Natural Language > Large Language Model (0.30)