TurboQuant: KV Cache Quantization to 3.5 Bits with Zero Accuracy Loss- ICLR 2026(darshanfofadiya.com)
darshanfofadiya.com
TurboQuant: KV Cache Quantization to 3.5 Bits with Zero Accuracy Loss- ICLR 2026
https://darshanfofadiya.com/research-papers/turboquant/